Task execution method and device based on index platform, equipment and medium

Abstract

The invention discloses a task execution method and device based on an index platform, equipment and a medium. The method comprises the steps that a to-be-executed task input by a user to an index platform is received through a preset intelligent dialogue interface, intention recognition is conducted on the to-be-executed task through a preset classification model, and a corresponding task label is obtained; calling a corresponding preset execution mode to execute the to-be-executed task according to the task label, and generating a corresponding task execution result; and displaying the task execution result through a preset display page. By implementing the method provided by the invention, the problem of poor application effect of the index platform in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a task execution method, device, equipment and medium based on an indicator platform. Background Art

[0002] With the rapid progress of big data and artificial intelligence technology, data indicator platforms have been widely used in the fields of enterprise management, business analysis and scientific research. These platforms can collect, process and analyze large amounts of data, providing decision makers with rich indicator information. However, faced with massive and complex indicator data, users are often confused and find it difficult to quickly locate key information or effectively interpret data. The indicator platforms currently on the market mainly use technical means such as traditional query screening, predefined reports, rule-based intelligent query and natural language processing. Traditional query screening operations are complex, predefined reports are not flexible enough, and rule-based queries are unable to cope with complex needs. Although the platform that initially applies large language model technology allows natural language query, it is limited by intent recognition and semantic understanding technology, and the accuracy and relevance of query results still need to be improved, so that the application effect of the indicator platform of existing technology still needs to be improved. Summary of the invention

[0003] The embodiments of the present invention provide a task execution method, device, equipment and medium based on an indicator platform, aiming to solve the problem of poor application effect of the indicator platform in the prior art.

[0004] In the first aspect, an embodiment of the present invention provides a task execution method based on an indicator platform, which includes: receiving a task to be executed input by a user to the indicator platform through a preset intelligent dialogue interface, identifying the intent of the task to be executed through a preset classification model, and obtaining a corresponding task label; calling a corresponding preset execution method according to the task label to execute the task to be executed, and generating a corresponding task execution result; and displaying the task execution result through a preset display page.

[0005] In the second aspect, an embodiment of the present invention also provides a task execution device based on an indicator platform, which includes: an identification unit, which is used to receive the tasks to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, and to identify the intention of the tasks to be executed through a preset classification model to obtain a corresponding task label; an execution unit, which is used to call the corresponding preset execution method according to the task label to execute the tasks to be executed and generate a corresponding task execution result; a display unit, which is used to display the task execution result through a preset display page.

[0006] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the computer program.

[0007] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, wherein the computer program includes program instructions, and the program instructions can implement the above method when executed by a processor.

[0008] The embodiment of the present invention provides a task execution method, device, equipment and medium based on an indicator platform. Among them, the method includes: receiving the task to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, identifying the intent of the task to be executed through a preset classification model, and obtaining the corresponding task label; calling the corresponding preset execution method according to the task label to execute the task to be executed, and generating the corresponding task execution result; displaying the task execution result through a preset display page. When the platform and the user have an intelligent dialogue, the embodiment of the present invention receives the task to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, and identifies the intent of the user's task to be executed, and calls the corresponding preset execution method according to the intention recognition result, so that whether it is retrieving specific indicator data, interpreting indicator data or drawing charts based on data, it can intelligently analyze user needs and provide accurate responses and results. Executing the task to be executed through the preset execution method and generating the corresponding task execution result can meet the user's various needs in a targeted manner. By performing intent recognition and calling the corresponding execution method to execute the task to be executed, the shortcomings of the existing technology in user experience, natural language understanding, data interpretation and data visualization are solved. It also significantly improves the intelligence level of the indicator platform, enhances flexibility and scalability, and can better adapt to user needs in complex business scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0010] Figure 1 A flowchart of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 2 A schematic diagram of a sub-process of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 3A schematic diagram of a sub-process of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 4 A schematic diagram of a sub-process of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 5 A schematic diagram of a sub-process of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 6 A schematic diagram of a sub-process of a task execution method based on an indicator platform provided in an embodiment of the present invention; Figure 7 A schematic block diagram of a task execution device based on an indicator platform provided by an embodiment of the present invention; Figure 8 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0013] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0014] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0015] See also Figure 1 , Figure 1A flowchart of a task execution method based on an indicator platform provided in an embodiment of the present invention. The task execution method based on an indicator platform in this embodiment can be applied to an indicator platform that can interact with users, wherein the indicator platform has an intelligent dialogue interface that can receive tasks conveyed by users and feed back processed results to users, and has an advanced large language model that can accurately identify users' tasks and their query intentions. By adopting the method of the present invention, the intelligence level of the indicator platform can be improved to meet the needs of users in complex business scenarios.

[0016] Figure 1 The flowchart of the task execution method based on the indicator platform provided by the embodiment of the present invention is shown in the figure. As shown in the figure, the method includes the following steps S110-S130.

[0017] S110, receiving the tasks to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, performing intent recognition on the tasks to be executed through a preset classification model, and obtaining corresponding task labels.

[0018] In this embodiment, the preset intelligent dialogue interface is a page for the platform to interact with the user. It can be a chat box on a web page, or other forms of interactive interfaces. Users input their needs through this interface to generate tasks to be executed on the indicator platform, such as querying the data of a certain indicator, updating indicator information, etc. The preset intelligent dialogue interface receives the tasks to be executed input by the user to the indicator platform. Specifically, when the user enters the task that needs to be executed by the platform in the intelligent dialogue interface, the platform will capture and receive these inputs. The input can be text, voice or other forms of data, depending on the design of the system and the user's choice, and there is no limitation on this. The preset classification model is a trained machine learning model for identifying the intention of the user input. In this embodiment, the preset classification model is a multi-classification model based on BERT (a model that uses a BERT (Bidirectional Encoder Representations from Transformers) pre-trained language model for multi-category classification tasks), wherein the preset classification model has learned and understood a variety of possible user intentions, and can match and identify according to the content input by the user. The preset classification model is used to identify the intention of the task to be executed, obtain the corresponding task label, and specifically, pass the task to be executed to the preset classification model. The classification model will analyze the input and determine the user's true intention. And determine the corresponding task label based on the intention recognition result, wherein the task label is a pre-set label corresponding to different execution methods, such as a data operation label and an intelligent analysis label. If the user's input is "I want to query last month's sales data", the classification model will recognize that the user's intention is "query sales data", and the recognized intention corresponds to the data operation label, then the preset classification model will generate the corresponding data operation label based on the task to be executed. By performing intent recognition on the task to be executed through the preset classification model and obtaining the corresponding task label, the user's intention can be accurately understood, and the correct task label can be assigned to each task to be executed, thereby ensuring that subsequent task execution and processing can proceed smoothly.

[0019] In one embodiment, if Figure 2 As shown, the step S110 also includes steps S1101-S1103 before the step S110.

[0020] S1101, generating a corresponding training data set according to a preset manual annotation or data filling method, wherein the data in the training data set are all provided with task labels; S1102, training the preset classification model using the preset written code and the training data set, and testing the trained preset classification model using a preset test set to obtain a test result; S1103: Adjust classification parameters in the preset classification model according to the test results.

[0021] In this embodiment, before the step of generating the corresponding training data set according to the preset manual annotation or data filling method, it includes: collecting raw data, performing data cleaning, data format conversion, data normalization and other pre-processing operations on the raw data, wherein data cleaning mainly removes duplicate data, processes missing values, corrects erroneous data, etc.; data format conversion converts data into the format required for model training; data normalization scales data into a unified range to improve the training efficiency and performance of the model. The operator annotates the pre-processed raw data according to the task label, and generates training data with task labels according to the preset filling rules, wherein the specific preset filling rules are not limited, and the corresponding data can be generated. After completing the manual annotation and data filling, the filled data is combined with the manually annotated data to construct the final training data set. The data in the training data set are all provided with task labels. Specifically, a file containing all possible labels is created for subsequent data annotation and model training. The file contains five labels: | (as a blank label or separator), chart, analysis, query, and export. These labels represent a blank (which may be used for no label or separation purposes) and a label corresponding to the execution mode. Each data in the training data set has one or more labels in the label file, indicating one or more task intentions that the data may involve, and the labels of the text data are converted into a numerical format that the model can understand. Specifically, for each label, a binary vector equal to the number of labels is created, where the position corresponding to the label is 1 and the remaining positions are 0. For example, if the label is chart, the vector may be [0, 1, 0, 0, 0] (assuming that the labels are arranged in order). In summary, the training data in the training data set has the task labels after format conversion. The preset classification model is trained with the training data set through the preset writing code. Specifically, the parameter information in the preset writing code is: set the training batch_size (the number of data processed simultaneously during each training) to 64, the epoch (the number of traversals of the training data set) to 10, the optimizer is Adam, the learn-rate (deep learning optimization algorithm) is 1e-5, and the base model is albert_small_zh_google (a pre-trained Chinese language model for feature extraction and initial model construction). After the preset classification model is trained, the trained preset classification model is tested using a preset test set, wherein the data in the preset test set does not have a task label. The preset test set is label-classified by the trained preset classification model to evaluate its generalization ability.The classification parameters in the preset classification model are adjusted according to the test results. Specifically, if the test results show that the preset classification model is underfitting or overfitting, the corresponding parameters used for classification (such as learning rate, batch size, number of epochs, etc.) are adjusted or the training set is re-prepared (such as increasing the amount of data, improving the quality of annotation, etc.). By adjusting the classification parameters in the preset classification model, the preset classification model can be continuously iterated and optimized, thereby constructing a classification model that can accurately identify natural language intent.

[0022] S120: Calling a corresponding preset execution method according to the task tag to execute the task to be executed, and generating a corresponding task execution result.

[0023] In this embodiment, the task label is a symbol or code for classifying tasks. Each task label corresponds to a specific type of task or operation. The preset execution method is a processing method or process preset for each task label. These execution methods can be program codes, algorithms, etc., which define how to perform corresponding operations for specific task labels. The task to be executed is executed by calling the corresponding preset execution method according to the task label. Specifically, when the task to be executed is received, the task label is first parsed to determine the type of task. Then, the corresponding preset execution method is searched according to the task label and called to process the task to be executed. After execution, the corresponding task execution result is generated. By calling the corresponding preset execution method according to the task label to execute the task and generate the task execution result, the platform can flexibly handle various types of tasks, thereby improving the system's automation level and user experience.

[0024] In one embodiment, if Figure 3 As shown, step S120 includes steps S121-S124.

[0025] S121: If the task tag is the data operation tag, analyzing the data operation type in the task to be executed by using a preset operation classification model; S122, performing data analysis on the task to be executed by using a preset offline large model to obtain a data operation target; S123, generating an operation statement according to the data operation type and the data operation target; S124. Execute the task to be executed according to the operation statement to generate a corresponding task execution result.

[0026] In this embodiment, the task label includes a data operation label, wherein the data operation label refers to a label for processing the data in the indicator platform, for example, operations such as adding, deleting, modifying and checking the data correspond to the data operation label. If the task label of the task to be executed is a data operation label, the data operation type in the task to be executed is analyzed by a preset operation classification model. Specifically, the preset operation classification model is a pre-trained machine learning or deep learning model for identifying and analyzing the data operation type in the task to be executed. In this embodiment, it is a multi-classification model based on BERT. Data operation types include operations such as insert, update, delete, and query. The task to be executed is analyzed by a preset offline large model to obtain the data operation target. Specifically, the preset offline large model is a large machine learning or deep learning model that is pre-trained and deployed in an offline environment, which is used to perform in-depth analysis of the data in the task to determine the target of the data operation. After determining the data operation type, the preset offline large model is used to analyze the data involved in the task and determine the specific target of the data operation. This includes identifying data records that need to be queried, inserted, updated or deleted, and determining the specific content of these records. An operation statement is generated according to the data operation type and the data operation target. Specifically, the operation statement is a database operation statement used to instruct the database system to perform the required data operation. The generated operation statement is executed by the database system to complete the task to be executed. The execution result will reflect the success or failure of the data operation and any related output data. For example, if the task to be executed is: "Update inventory quantity", the data operation type is Update, and a preset offline large model is used to analyze the data involved in the task. It identifies the inventory records that need to be updated and determines the new inventory quantity, thereby generating an SQL update statement, such as: "UPDATE inventory SET quantity = 100 WHERE product_id = 123". This statement instructs the database system to update the inventory quantity of a specific product. By generating an operation statement according to the data operation type and the data operation target, the conversion from natural language to database language is realized, and the indicator data required by the user can be quickly and accurately retrieved from massive data. Compared with the traditional manual query method, the data retrieval speed and accuracy have been significantly improved.

[0027] In addition, in this embodiment, before the step of analyzing the data operation type in the task to be executed by the preset operation classification model, it also includes: constructing an instruction data set, wherein the instruction data set should include at least one operation instruction of adding, deleting, modifying and checking, converting the instruction data set into a format, and constructing an indicator operation training set, for example, creating a One-Hot encoding for each operation instruction. For example, "add" may be encoded as [1, 0, 0, 0], "delete" as [0, 1,0, 0], "modify" as [0, 0, 1, 0], and "query" as [0, 0, 0, 1]. And train the preset operation classification model according to the indicator operation training set, and perform SFT (Supervised Fine-Tuning) fine-tuning training on the offline large model. And adjust the training parameters such as the learning rate of the model according to the training results. The accuracy of the model can be improved by training the preset operation classification model and the offline large model.

[0028] In one embodiment, if Figure 4 As shown, step S120 includes steps S125-S127.

[0029] S125. If the task tag is the data visualization tag, determine whether there is a field in the task to be executed that is consistent with the preset visualization graphic name according to a preset matching rule; S126, if it does not exist, performing similarity matching between the task to be executed and the preset visualization graphic name to obtain a corresponding visualization type; S127. Call a corresponding visualization chart generation function according to the visualization type to generate a corresponding visualization result.

[0030] In this embodiment, the task label includes a data visualization label, and the data visualization refers to the representation of data in the form of graphics or the like. If the task label of the task to be executed is a data visualization label, it is determined whether there is a field in the task to be executed that is consistent with the preset visualization graphic name according to the preset matching rules. Specifically, the field in the task to be executed is checked according to the preset matching rules to determine whether there is a field that is consistent with the preset visualization graphic name. Among them, the preset matching rules can be rules such as matching based on field names and regular expressions, and the specific rules are not limited. If there is no field in the task to be executed that is completely consistent with the preset visualization graphic name, a similarity matching algorithm will be used to find the most matching visualization graphic name, wherein the similarity matching method includes cosine similarity, Jaccard similarity, Levenshtein distance, etc. These algorithms are used to calculate the similarity between the field in the task to be executed and the preset visualization graphic name, wherein the specific algorithm used is not limited. Based on the result of the similarity matching, the system will select the visualization graphic name that best matches the task to be executed and determine the corresponding visualization type. For example, if the task to be executed involves the name of the visualization graphic "main installation diagram", it cannot be completely consistent with the preset visualization graphic name. Therefore, a similarity matching algorithm (such as cosine similarity) is used to calculate the similarity between the field value and the "bar chart" and determine that they are the best match. Then the visualization type is determined to be a bar chart. Then the function for generating a bar chart is called, and the task target carried in the task to be executed (such as sales data for the first quarter of 2023) is passed in to generate the required bar chart visualization result. By automatically identifying and generating corresponding charts (such as bar charts, pie charts, line charts, etc.) according to the user's natural language request to intuitively and accurately display the indicator data in the indicator platform, the visualization effect of the data is improved.

[0031] In one embodiment, if Figure 5 As shown, the step S120 includes steps S128-S1210.

[0032] S128. If the task tag is the file export tag, determining whether the to-be-executed task has a task export format that is the same as the preset export format according to a preset data matching rule; S129, if the target export format does not exist, matching the task to be executed with the preset export format for similarity to determine the target export format; S1210. Call a corresponding data export function according to the target export format to generate a corresponding export file.

[0033] In this embodiment, the task tag includes a file export tag, which refers to a tag that exports part of the indicator data in a certain format file. If the task tag is the file export tag, it is determined whether the task to be executed has a task export format that is the same as the preset export format according to the preset data matching rule, specifically, it is determined whether the export format of the task to be executed matches one of the preset export formats. The preset export format is a set of export formats predefined by the system to meet different user needs or business requirements. If the export format of the task to be executed is the same as one of the preset export formats, the preset export format can be used directly; if there is no matching export format, the similarity matching algorithm is used to find the preset export format that is most similar to the export format of the task to be executed as the target export format. After the target export format is determined, the data export function corresponding to the target export format is called. This function is responsible for generating the final export file according to the data content of the task to be executed and the requirements of the target export format. Among them, the preset export format and the similarity matching method are similar to the preset matching rules and similarity matching methods of the above-mentioned data visualization tags, which will not be repeated. By calling the corresponding data export function according to the target export format, the indicator platform can flexibly handle the export task requested by the user, even if the export format requested by the user is not one of the system preset formats.

[0034] In one embodiment, the step S120 includes step S1211.

[0035] S1211. If the task tag is the intelligent analysis tag, data analysis is performed on the analysis requirements and indicator data carried in the task to be executed according to a preset fusion analysis model to generate corresponding data analysis results.

[0036] In this embodiment, the task tag includes an intelligent analysis tag. If it is an intelligent analysis tag, it means that the task to be performed is to analyze certain indicator data. The preset fusion analysis model refers to combining multiple analysis techniques and methods to achieve comprehensive and in-depth analysis of complex data. This model usually combines multiple technologies such as data mining, data abstraction, association rule analysis, decision tree analysis, etc. In this embodiment, the preset fusion analysis model is a large model after training. According to the preset fusion analysis model, the analysis requirements and indicator data carried in the task to be performed are analyzed. Specifically, the analysis requirements refer to the explicit or implicit data analysis requirements in the task to be performed. These requirements may involve aspects such as the type, quantity, quality, and analysis purpose of the data. The indicator data refers to data used to analyze and evaluate relevant factors or variables in the task to be performed. According to the preset fusion analysis model, the analysis requirements and indicator data in the task to be performed are processed. Finally, the corresponding data analysis results are generated, which are usually presented in charts, reports, or other visual forms for users to understand and use. It is understandable that a task to be executed will include one or more task tags. When there are multiple task tags, all results need to be integrated to obtain the corresponding task analysis results. For example, the data can be analyzed first, and a bar chart can be generated based on the analyzed data. Among them, there is no limitation on the specific integration method and order, as long as the corresponding results can be fully displayed. By performing data analysis on the analysis requirements and indicator data carried in the task to be executed according to the preset fusion analysis model, the corresponding data analysis results are generated to help users understand the phenomena and trends behind the data.

[0037] In one embodiment, if Figure 6 As shown, the step S1211 includes steps S12111-S12113.

[0038] S12111. Fine-tune the sub-models of the base model according to the first fusion training set, adjust parameters of the trained preset fusion analysis model, and fuse information of the sub-model with the best performance; S12112, using the second fusion training set to perform reinforcement training on the sub-model after information fusion by a preset reinforcement learning method, and adjusting parameters of the trained sub-model; S12113. According to the enhanced training effect, the parameter weight information of the sub-model with the best performance is integrated to determine the preset fusion analysis model.

[0039] In this embodiment, the first fusion training set is a set of analysis reports for fine-tuning the sub-model. Specifically, the corresponding analysis report can be generated according to the manually written part of the indicator data, and a batch of analysis reports can be generated through large models such as Wenxin Yiyan. The first fusion training set is determined jointly based on the manually written reports and the analysis reports generated by the large model. The base model is a discrete large model. Specifically, the base model is constructed by decomposing a larger model into multiple sub-models, each of which focuses on solving a specific sub-task. The prediction results of these sub-models are combined to form the final prediction result. The sub-models of the base model are fine-tuned according to the first fusion training set, and the parameters of the trained sub-models are adjusted, and the sub-model with the best performance is subjected to information fusion. Specifically, the fine-tuning training is to keep most of the model parameters unchanged, and train some parts of the model (such as the last layer) to adapt to new data or tasks. For example, assuming that the base model has four sub-models A, B, C, and D, they are first fine-tuned using the first fusion training set. After fine-tuning, it is found that models A, B, and D perform best on specific tasks. Then, A, B and D are fused together using a certain fusion strategy (such as weighted average). The second fusion training set is a data set for reinforcement learning, and the construction method can be the same as the first fusion training set. The preset reinforcement learning method is a machine learning method in which the model learns how to take the best action to maximize a certain reward by interacting with the environment. For example, such as Q-learning or deep reinforcement learning. The model that has been fused with A, B and D is trained by the second fusion training set and the preset reinforcement learning method (such as deep reinforcement learning). During the training process, the model parameters are adjusted according to the performance of the model on the task to maximize the reward (such as prediction accuracy). The reinforcement training effect refers to the performance of the model during the reinforcement training process, which is usually measured by indicators such as reward or accuracy. After reinforcement training, the best performing sub-model is selected and its parameter weight information is fused into the base model. By fusing the parameter weight information of the optimal sub-model, a fusion analysis model with better performance is determined. For example, after the reinforcement training is completed, it is assumed that some parameter weights of model A are found to perform particularly well on a specific task. Then, these weight information can be fused into the base model to form a preset fusion analysis model with better performance. The model is trained in multiple aspects such as fine-tuning, fusion, reinforcement training and parameter adjustment to improve the performance of the model through continuous iteration and optimization.

[0040] S130: Displaying the task execution result through a preset display page.

[0041] In this embodiment, the preset display page refers to a user interface (UI) pre-designed by the system for displaying the results of task execution. These pages usually include various elements, such as text, images, charts, buttons, links, etc., which are used to present information, receive user input, and guide user operations. The task execution result refers to the output generated after the platform executes the task, such as data analysis results, information query results, etc. By displaying the task execution results in the preset display page, an intuitive, clear, and easy-to-use interface can be provided to the user, thereby improving the user experience and satisfaction of the system.

[0042] Figure 7 is a schematic block diagram of a task execution device 200 based on an indicator platform provided by an embodiment of the present invention. Figure 7 As shown, corresponding to the above-mentioned task execution method based on the indicator platform, the present invention also provides a task execution device based on the indicator platform. The task execution device based on the indicator platform includes a unit for executing the above-mentioned task execution method based on the indicator platform, and the device can be configured in a desktop computer, a tablet computer, a laptop computer, etc. Specifically, please refer to Figure 7 The task execution device based on the indicator platform includes an identification unit 210, an execution unit 220, and a display unit 230.

[0043] The identification unit 210 is used to receive the tasks to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, perform intent identification on the tasks to be executed through a preset classification model, and obtain corresponding task labels.

[0044] In one embodiment, the identification unit 210 includes a filling unit, a testing unit, and an adjustment unit.

[0045] A filling unit, used to generate a corresponding training data set according to a preset manual annotation or data filling method, wherein the data in the training data set are all provided with task labels; A testing unit, used to train the preset classification model by using a preset code and the training data set, and to test the trained preset classification model using a preset test set to obtain a test result; An adjustment unit is used to adjust the classification parameters in the preset classification model according to the test result.

[0046] The execution unit 220 is used to call the corresponding preset execution mode according to the task tag to execute the task to be executed and generate a corresponding task execution result.

[0047] In one embodiment, the execution unit 220 includes an analysis unit, an acquisition unit, a generation unit, and an execution sub-unit.

[0048] an analyzing unit, configured to analyze the data operation type in the task to be executed by using a preset operation classification model if the task tag is the data operation tag; An acquisition unit, used to perform data analysis on the task to be executed through a preset offline large model to acquire a data operation target; A generating unit, configured to generate an operation statement according to the data operation type and the data operation target; The execution subunit is used to execute the task to be executed according to the operation statement to generate a corresponding task execution result.

[0049] In one embodiment, the execution unit 220 includes a first judgment unit, a first matching unit, and a first calling unit.

[0050] A first judgment unit is used to judge whether there is a field in the to-be-executed task that is consistent with the preset visualization graphic name according to a preset matching rule if the task tag is the data visualization tag; A first matching unit, configured to perform similarity matching between the task to be executed and the preset visualization graphic name if the task does not exist, to obtain a corresponding visualization type; The first calling unit is used to call the corresponding visualization chart generation function according to the visualization type to generate a corresponding visualization result.

[0051] In one embodiment, the execution unit 220 includes a second judgment unit, a second matching unit, and a second calling unit.

[0052] A second judgment unit is used for judging whether the to-be-executed task has a task export format that is the same as a preset export format according to a preset data matching rule if the task tag is the file export tag; A second matching unit, configured to perform similarity matching between the task to be executed and the preset export format if the target export format does not exist, to determine the target export format; The second calling unit is used to call the corresponding data export function according to the target export format to generate a corresponding export file.

[0053] In one embodiment, the execution unit 220 includes a data analysis unit.

[0054] A data analysis unit is used to perform data analysis on the analysis requirements and indicator data carried in the task to be executed according to a preset fusion analysis model if the task tag is the intelligent analysis tag, and generate corresponding data analysis results.

[0055] In one embodiment, the execution unit 220 includes a fine-tuning unit, an enhancement unit, and a fusion unit.

[0056] A fine-tuning unit, used to perform fine-tuning training on the sub-model of the base model according to the first fusion training set, adjust parameters of the trained preset fusion analysis model, and perform information fusion on the sub-model with the best performance; A reinforcement unit, used for performing reinforcement training on the sub-model after information fusion using the second fusion training set through a preset reinforcement learning method, and adjusting parameters of the trained sub-model; The fusion unit is used to fuse the parameter weight information of the sub-model with the best performance according to the enhanced training effect to determine the preset fusion analysis model.

[0057] The display unit 230 is used to display the task execution result through a preset display page.

[0058] It should be noted that technical personnel in the relevant field can clearly understand that the specific implementation process of the above-mentioned task execution device 200 based on the indicator platform and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and conciseness of the description, it will not be repeated here.

[0059] The task execution device based on the indicator platform can be implemented in the form of a computer program. Figure 8 Runs on the computer device shown.

[0060] See also Figure 8 , Figure 8 5 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a terminal or a server, wherein the terminal may be an electronic device with communication functions such as a smart phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The server may be an independent server or a server cluster composed of multiple servers.

[0061] See also Figure 8 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0062] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions, and when the program instructions are executed, the processor 502 can execute a task execution method based on an indicator platform.

[0063] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0064] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a task execution method based on an indicator platform.

[0065] The network interface 505 is used to communicate with other devices over the network. Figure 8 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0066] The processor 502 is used to run a computer program 5032 stored in the memory to implement the steps of the above method.

[0067] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0068] It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0069] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the processor executes the steps of the above method.

[0070] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0071] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0072] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0073] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0074] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0075] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A task execution method based on an indicator platform, characterized in that: include: Receiving the tasks to be executed input by the user to the indicator platform through a preset intelligent dialogue interface, performing intent recognition on the tasks to be executed through a preset classification model, and obtaining corresponding task labels; Calling a corresponding preset execution method according to the task tag to execute the task to be executed, and generating a corresponding task execution result; The task execution results are displayed through a preset display page.

2. The method according to claim 1, characterized in that Before the step of identifying the intent of the task to be executed by using a preset classification model and obtaining the corresponding task label, the method includes: Generate a corresponding training data set according to a preset manual annotation or data filling method, wherein the data in the training data set are all provided with task labels; The preset classification model is trained by using the preset code and the training data set, and the trained preset classification model is tested using the preset test set to obtain the test result; The classification parameters in the preset classification model are adjusted according to the test results.

3. The method according to claim 1, characterized in that: The task tag includes a data operation tag, and the step of calling a corresponding preset execution mode according to the task tag to execute the task to be executed and generating a corresponding task execution result includes: If the task tag is the data operation tag, analyzing the data operation type in the task to be executed by using a preset operation classification model; Performing data analysis on the task to be executed by using a preset offline large model to obtain a data operation target; Generate an operation statement according to the data operation type and the data operation target; The task to be executed is executed according to the operation statement to generate a corresponding task execution result.

4. The method according to claim 1, characterized in that The task tag includes a data visualization tag, and the step of calling a corresponding preset execution mode according to the task tag to execute the task to be executed and generating a corresponding task execution result includes: If the task tag is the data visualization tag, determining whether there is a field in the task to be executed that is consistent with the preset visualization graphic name according to a preset matching rule; If it does not exist, the task to be executed is matched with the preset visualization graphic name by similarity to obtain the corresponding visualization type; The corresponding visualization chart generation function is called according to the visualization type to generate a corresponding visualization result.

5. The method according to claim 1, characterized in that The task tag includes a file export tag, and the step of calling a corresponding preset execution mode according to the task tag to execute the task to be executed and generating a corresponding task execution result includes: If the task tag is the file export tag, determining whether the to-be-executed task has a task export format that is the same as the preset export format according to a preset data matching rule; If it does not exist, the task to be executed is matched with the preset export format for similarity to determine the target export format; The corresponding data export function is called according to the target export format to generate a corresponding export file.

6. The method according to claim 1, characterized in that The task tag includes an intelligent analysis tag, and the step of calling a corresponding preset execution mode according to the task tag to execute the task to be executed and generating a corresponding task execution result includes: If the task tag is the intelligent analysis tag, data analysis is performed on the analysis requirements and indicator data carried in the task to be executed according to a preset fusion analysis model to generate corresponding data analysis results.

7. The method according to claim 6, characterized in that Before the step of performing data analysis on the analysis requirements and indicator data carried in the task to be executed according to the preset fusion analysis model, the step includes: Fine-tune the sub-models of the base model according to the first fusion training set, adjust the parameters of the trained preset fusion analysis model, and fuse the information of the sub-model with the best performance; Using the second fusion training set to perform reinforcement training on the sub-model after information fusion through a preset reinforcement learning method, and adjusting parameters of the trained sub-model; According to the enhanced training effect, the parameter weight information of the sub-model with the best performance is fused to determine the preset fusion analysis model.

8. A task execution device based on an indicator platform, characterized in that: include: An identification unit, used to receive a task to be executed input by a user to the indicator platform through a preset intelligent dialogue interface, perform intent identification on the task to be executed through a preset classification model, and obtain a corresponding task label; An execution unit, configured to execute the task to be executed by calling a corresponding preset execution mode according to the task tag, and generate a corresponding task execution result; The display unit is used to display the task execution result through a preset display page.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the method according to any one of claims 1 to 7 can be implemented.

Images