A platform reporting method, system and medium based on AIGC robot

Abstract

The present invention discloses a platform reporting method, system and medium based on AIGC robot, and relates to the field of artificial intelligence technology. The method includes: after creating the AIGC robot, defining its role logic and task objectives according to the requirements, and reading the creator's identity and authority information to complete the initialization; using the initialized robot to talk with the creator and establish an adaptive feedback layer; calling the decision layer to perform reply decision authentication and generate authentication results; when the result is an output decision, calling the multimodal data analysis network to perform multidimensional business data integration and generate integration results; optimizing the integration results according to the requirements to generate an output report, obtain the creator's feedback, and use the feedback to update the adaptive feedback layer to solve the technical problems that the existing artificial intelligence platform based on the large language model is insufficient in the depth of context understanding, real-time and accuracy of generated content.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a platform reporting method, system and medium based on an AIGC robot. Background Art

[0002] With the rapid development of artificial intelligence technology, generative artificial intelligence (AIGC) based on large language models has gradually become a key tool in the field of intelligent interaction and data processing. With its powerful language generation ability and breadth of knowledge coverage, this type of technology has shown significant advantages in scenarios such as intelligent customer service and content creation. However, in the field of complex business needs and accurate report generation, existing technologies still have many shortcomings. In the long-term dialogue with users, the current large language model often finds it difficult to achieve deep contextual understanding and semantic coherence, and is prone to inaccurate answers due to information loss or misunderstanding. In addition, the training data of these models is usually based on a specific time point and cannot update new knowledge in a timely manner, especially when it comes to real-time events or dynamic business needs. There are also problems with the accuracy and reliability of the generated content. For example, in scenarios where high-precision business data integration and analysis are required, the model may provide incorrect or incomplete answers, affecting user decisions. At the same time, existing technologies mainly output results in plain text and lack intuitive data visualization capabilities. This is particularly insufficient for complex scenarios that require multi-dimensional business data integration, statistical analysis, and dynamic report generation. Especially when users need to call data across platforms and generate reports combining graphics and text, the limitations of traditional systems become more apparent and cannot meet efficient, intelligent, and visual business needs. Summary of the invention

[0003] The present application provides a platform reporting method, system and medium based on an AIGC robot, which solves the technical problems of the existing artificial intelligence platform based on a large language model in terms of insufficient depth of context understanding, real-time performance and accuracy of generated content. By introducing identity authority information reading, an adaptive feedback layer and a multimodal data analysis network, a task-specific role logic and a multi-dimensional business data integration mechanism are established, achieving the technical effects of enhancing the depth of context understanding, improving the accuracy of data generation and optimizing the user interaction experience.

[0004] In view of the above problems, the present application provides a platform reporting method, system and medium based on an AIGC robot.

[0005] In a first aspect of the present application, a platform reporting method based on an AIGC robot is provided, the method comprising: after creating the AIGC robot, defining the role logic and task objectives of the AIGC robot according to the needs of the creator, reading the identity information and permission information of the creator, and initializing the AIGC robot with the identity information and the permission information; using the initialized AIGC robot to conduct a dialogue with the creator, and establishing an adaptive feedback layer, the adaptive feedback layer being used to record the operating characteristics and demand characteristics of the creator; calling the decision layer of the AIGC robot to perform a reply decision authentication of the dialogue, and establishing a reply decision authentication result; when the reply decision authentication result is an output result decision, calling a multimodal data analysis network according to the output result decision to perform multidimensional business data integration; and establish a multi-dimensional business data integration result; after optimizing the multi-dimensional business data integration result according to the demand characteristics, generate an output report, obtain creator feedback, and use the creator feedback to update the adaptive feedback layer; wherein, calling the decision layer of the AIGC robot to perform a reply decision authentication for the dialogue and establish a reply decision authentication result includes: using the decision layer to perform dialogue backtracking and establish a backtracking data set; obtaining the dialogue features of the current round of dialogue, using the dialogue features to perform a time series correlation analysis of the backtracking data set, and using the time series correlation results to create a context window, the context window having a context trust mark; performing a reply decision value calculation for the current round of dialogue in the context window to generate a reply decision value calculation result; and completing the reply decision authentication after compensating the reply decision value calculation result by the context trust mark.

[0006] The second aspect of the present application provides a platform reporting system based on an AIGC robot, the system comprising:

[0007] Robot initialization processing component: after creating the AIGC robot, define the role logic and task objectives of the AIGC robot according to the creator's needs, read the identity information and permission information of the creator, and use the identity information and permission information to initialize the AIGC robot; adaptive feedback layer establishment component: use the initialized AIGC robot to conduct a dialogue with the creator, and establish an adaptive feedback layer, the adaptive feedback layer is used to record the creator's operating characteristics and demand characteristics; reply decision authentication component: call the decision layer of the AIGC robot to perform reply decision authentication of the dialogue, and establish a reply decision authentication result; multi-dimensional business data integration component: when the reply decision authentication result is an output result decision, call the multimodal data analysis network according to the output result decision to perform multi-dimensional business data integration, and establish a multi-dimensional business data integration result; output report generation component: after optimizing the multi-dimensional business data integration result according to the demand characteristics, generate an output report, obtain creator feedback, and use the creator feedback to update the adaptive feedback layer.

[0008] The third aspect of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a platform reporting method based on an AIGC robot provided in the present application.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] First, after creating the AIGC robot, the role logic and task objectives of the AIGC robot are defined according to the needs of the creator, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and the permission information. Subsequently, the initialized AIGC robot is used to conduct a dialogue with the creator, and an adaptive feedback layer is established, which is used to record the operating characteristics and demand characteristics of the creator. Further, the decision layer of the AIGC robot is called to perform the reply decision authentication of the dialogue, and the decision authentication result of the reply is established. Afterwards, when the decision authentication result of the reply is an output result decision, the multimodal data analysis network is called according to the output result decision to perform multidimensional business data integration, and a multidimensional business data integration result is established. Finally, after optimizing the multidimensional business data integration result according to the demand characteristics, an output report is generated, and the creator feedback is obtained, and the adaptive feedback layer is updated using the creator feedback. It solves the technical problems of insufficient context understanding depth, real-time performance and accuracy of generated content in existing artificial intelligence platforms based on large language models. By introducing identity and authority information reading, adaptive feedback layer and multimodal data analysis network, it establishes task-specific role logic and multi-dimensional business data integration mechanism, achieving the technical effect of enhancing the depth of context understanding, improving data generation accuracy and optimizing user interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flowchart of a platform reporting method based on an AIGC robot is provided in an embodiment of the present application.

[0013] Figure 2 A schematic diagram of the structure of a platform reporting system based on an AIGC robot provided in an embodiment of the present application.

[0014] Explanation of the reference numerals: robot initialization processing component 11, adaptive feedback layer establishment component 12, reply decision authentication component 13, multi-dimensional business data integration component 14, output report generation component 15. DETAILED DESCRIPTION

[0015] This application solves the technical problems of insufficient context understanding depth, real-time performance and accuracy of generated content in existing artificial intelligence platforms based on large language models by providing a platform reporting method, system and medium based on AIGC robots. By introducing identity authority information reading, adaptive feedback layer and multimodal data analysis network, task-specific role logic and multi-dimensional business data integration mechanism are established, achieving the technical effects of enhancing the depth of context understanding, improving data generation accuracy and optimizing user interaction experience.

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0017] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0018] Embodiment 1, as Figure 1 As shown, the present application provides a platform reporting method based on an AIGC robot, wherein the method comprises:

[0019] After the AIGC robot is created, the role logic and task objectives of the AIGC robot are defined according to the creator's needs, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and the permission information.

[0020] In one embodiment, the creator creates an AIGC robot through the button platform, which is usually done on the button platform's operating interface. The creator needs to click a specified button or select a corresponding option to generate a new robot instance. During the creation process, the button platform will assign a unique identity to the AIGC robot, and can select the robot's functional modules and basic configurations according to user needs. After the AIGC robot is created, the system terminal defines the role logic and task objectives of the AIGC robot according to the creator's needs, which means setting the AIGC robot's behavior mode and target tasks according to the application scenario and purpose of use of the AIGC robot, such as a fleet status monitoring assistant, a risk report generation assistant, etc. At the same time, during the initialization process, the platform The creator's identity information and permissions will be read to ensure that the robot provides different levels of services according to the identities and permissions of different creators in subsequent interactions. For example, if the creator is the administrator of the fleet management platform, the AIGC robot may have more operating permissions, such as access to all fleet data and the ability to generate global fleet status reports, while ordinary fleet users are subject to certain restrictions, such as only being able to view the status and alarm information of specific vehicles. This step is a prerequisite for the AIGC robot to be able to work flexibly according to the different needs and permissions of the creator. In summary, after the AIGC robot is created, the platform will personalize the robot initialization based on the creator's needs, identity, and permissions to ensure that the robot can perform specified tasks and interact effectively.

[0021] The initialized AIGC robot is used to conduct a dialogue with the creator and establish an adaptive feedback layer, wherein the adaptive feedback layer is used to record the operating characteristics and demand characteristics of the creator.

[0022] In one embodiment, after the creation of the initialized AIGC robot is completed, the AIGC robot begins to have actual dialogues and interactions with the creator. Through these dialogues, an adaptive feedback layer is gradually established. The core function of the adaptive feedback layer is to continuously record the creator's operational characteristics and demand characteristics in the interaction with the robot. For example, the AIGC robot can record the creator's commonly used query instructions, preferred report formats, or specific data of interest (such as fleet status, risk reports, etc.). This information is not only the content of the dialogue itself, but also includes the creator's usage mode, interaction habits, and behavioral preferences for the robot. By analyzing this data, the AIGC robot can continuously adjust its response and behavior to make it more accurate feedback to the creator's needs. In the specific application scenario of the fleet management platform, if the creator frequently queries a certain type of data or generates a certain type of report (such as a fleet status report), the AIGC robot will automatically give priority to these needs in subsequent dialogues and quickly provide relevant information without the need to re-enter each time. The adaptive feedback layer accumulates these operational and demand characteristics to optimize the AIGC robot's answer logic and response method, thereby improving interaction efficiency and accuracy. Ultimately, the robot can flexibly adjust service strategies according to the different needs and changes of the creator and provide more personalized services.

[0023] For the adaptive feedback layer, its specific construction process is as follows: In the initial stage, the AIGC robot starts to talk with the creator, records each interaction and its related operations, and collects the creator's operation characteristics and demand characteristics through the platform's dialogue interface; in terms of operation characteristic collection, the system terminal records the creator's input commands, query types, query frequencies, and common functions and other behavioral data, which will help understand the creator's interaction habits and preferences. At the same time, the creator's demand characteristics will be collected, such as the type of data, service requirements, report content, data display format, and query preferences that the creator is concerned about in the conversation. The collection of these demand characteristics will provide key support for the establishment of the adaptive feedback layer. For example, when the creator requests to generate a pie-shaped fleet status report for the first time, the pie-shaped fleet status report will be marked as a demand feature; when building, taking the fully connected neural network as an example, other algorithm models can also be used, such as random forests, policy models, recurrent neural networks, etc. The system terminal first needs to convert the collected features into a format suitable for neural network processing. For discrete categories (such as query types and common functions), one-hot encoding is used to convert them into vector representations. For numerical features (such as query frequency), the system terminal uses a unique encoding method to convert them into vector representations. rate), the original value can be used directly, or it can be normalized; then, a fully connected neural network structure is constructed, including an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer is equal to the dimension of the operation feature, that is, the number of all numerical features and encoded discrete features. The number of hidden layers and the scale of neurons can be adjusted according to actual conditions. The number of neurons in the output layer is equal to the number of categories of demand features; during the training process, the system terminal will input the creator's operation features into the fully connected neural network, and the network will predict the corresponding demand features through forward propagation. The gap between the predicted results and the actual demand features is calculated through the cross entropy loss function, and the weights in the network are adjusted through the Adam optimizer to minimize the loss. After each training cycle, the model will be adjusted according to the performance of the validation set to avoid overfitting and ensure that the fully connected neural network can stably predict demand features; after the training is completed, the fully connected neural network that meets the expected expectations will be used as the final adaptive feedback layer. This adaptive feedback layer will continue to be updated and optimized as the use time of the AIGC robot increases, ensuring that the AIGC robot can always efficiently and accurately meet the creator's needs in long-term interactions.

[0024] The decision layer of the AIGC robot is called to perform reply decision authentication for the dialogue, and establish a reply decision authentication result.

[0025] In one embodiment, when conversing with the creator, the AIGC robot needs to make an appropriate response based on the content of the conversation. To ensure the accuracy and expectedness of the response, the system terminal will call the decision layer of the AIGC robot to perform decision authentication on the AIGC robot's response. Specifically, the decision layer will analyze the current conversation content based on the context information of the conversation and the preset rules of the AIGC robot to determine whether the robot's response is reasonable and accurate. Based on the content, context and needs of the conversation, the decision layer will evaluate whether the answer generated by the AIGC robot meets the creator's expectations by calculating the response decision value, and generate a decision authentication result based on predetermined standards. This process ensures that the AIGC robot does not just generate a simple answer, but has undergone a layer of verification to ensure that the answer is logical, accurate and relevant to the context.

[0026] Furthermore, the calling of the decision layer of the AIGC robot to perform a reply decision authentication of the dialogue and establish a reply decision authentication result includes:

[0027] The decision layer is used to perform dialogue backtracking and establish a backtracking data set; the dialogue features of the current dialogue are obtained, the dialogue features are used to perform time series correlation analysis of the backtracking data set, and the time series correlation results are used to create a context window, wherein the context window has a context trust mark; the reply decision value of the current dialogue is calculated within the context window to generate a reply decision value calculation result; after compensating the reply decision value calculation result through the context trust mark, the reply decision authentication is completed.

[0028] Preferably, the system terminal first retraces the previous conversation content through the decision layer of the AIGC robot, and compiles a retracing data set, which contains key information in the past conversation, including the creator's inquiry, the response given by the AIGC robot, and the user's feedback. This information will serve as the basis for subsequent decision-making and analysis; then, extract the features of the current round of conversation, such as the topic, emotion, intention, etc. of the conversation, and perform temporal correlation analysis on the current conversation features and the retracing data set, and compensate the analysis results with weakened correlation factors, so that the system terminal can understand the correlation between the current conversation and the historical conversation, help identify the context changes of the conversation, and help it understand the position and importance of the current conversation in the entire conversation history; after completing the temporal correlation analysis, the robot creates a context window based on the analysis results. The window contains not only the information of the current conversation, but also the historical context data related to it. Through this information, the background and goal of the current conversation can be fully understood. The context window will also have a context trust mark, indicating the current context. Credibility is determined based on the confidence in the time-series correlation results. If the historical conversation is highly matched with the current conversation and the context trust indicator is high, the AIGC robot can make more accurate decisions. After that, the system terminal calculates the reply decision value in the context window, that is, it evaluates what type of answer should be given to the current conversation. The reply decision value is calculated based on the content of the context window, the creator's needs, and the characteristics of the conversation, indicating the rationality and relevance of the answer. Then, in order to make the calculation result more credible, the context trust indicator is used to compensate the reply decision value calculation result through weighted correction, so that the answer generated by the AIGC robot is more in line with the creator's needs and context. Finally, the compensated reply decision value is used to generate the final answer and complete the reply decision certification.

[0029] Furthermore, the using the conversation features to perform temporal correlation analysis of the retrospective data set includes:

[0030] The data timestamp of the retrospective data set is obtained, and a weakened correlation factor is constructed using the data timestamp; an association rule algorithm is used to perform an association analysis between the retrospective data set and the user emotion, user behavior and user intention of the conversation feature, and an association analysis result is established; and the association analysis result is compensated based on the weakened correlation factor to complete the time series association analysis.

[0031] Optionally, in the process of processing a conversation, the system terminal first obtains the timestamp of each historical conversation from the retrospective data set, that is, records the time when each conversation occurred. Through these timestamps, the time difference between the historical conversation and the current conversation can be understood. The system terminal constructs a weakened correlation factor in the form of exponential book decay based on the time difference between the current conversation and the historical conversation. Specifically, , where λ is the decay coefficient, which is used to control the decay speed and is determined based on actual needs and expert advice. Δt is the time difference between the current conversation and the historical conversation. In this way, a lower weight can be given to the conversation history with a longer time span, so that the closer the historical record is to the current conversation, the stronger its confidence is, and the older the conversation is, the weaker its impact on the current conversation will be. Subsequently, the association rule algorithm (such as Apriori algorithm, FP-growth, etc.) is used to analyze the retrospective dataset to find the relationship between the current conversation features (such as user emotions, behaviors, and intentions) and historical conversations. By analyzing user emotions (such as positive or negative emotions), It can identify the potential connection between the current conversation and the historical conversation, and construct a correlation analysis result. This result helps the robot understand the needs behind the current conversation and match it with historical data. For correlation analysis, taking FP-growth as an example, the feature data (emotion, behavior, intention) of the historical conversation and the current conversation is converted into a format suitable for algorithm analysis. The FP-growth algorithm usually needs to input a transaction data set, where each transaction represents a conversation, and the elements of the transaction are the features in the conversation, such as [emotion, behavior, Intent] format, and then take the features of each conversation as the elements of the item set, and count the frequency of each item set; then create an FP tree, and build a tree structure by traversing all conversation data. Each node of the tree represents a feature item, and the frequency of the node represents the number of times the feature appears in the data set. The tree construction process will sort the item sets from high to low according to their frequencies, ensuring that frequently occurring items are concentrated at the top of the tree; then extract frequent item sets from the FP tree, that is, combinations of feature items (such as emotions, behaviors, and intentions) in the conversation. These combinations occur frequently in the conversation, indicating that there is a strong correlation between them. For example, if the frequency of positive emotions and querying the fleet status appearing at the same time is greater than the preset frequent frequency, then the frequent item sets will be extracted from the FP tree. rate, the combination of these two features is taken as a frequent item set, indicating that the creator often queries the fleet status under positive emotions; based on frequent item sets, FP-growth can generate association rules in the form of {A}→{B}, which means that when A (such as positive emotions) occurs, B (such as querying the fleet status) is also more likely to occur. The confidence of each rule represents the probability of B occurring when A occurs, which is calculated by comparing the frequency of A and B occurring at the same time with the frequency of A occurring; then, the historical conversations corresponding to the confidence greater than the preset confidence are extracted and used as the output results of the association analysis together with their respective confidences to establish the association analysis results;Since the time difference of historical conversations may affect the understanding of the current conversation, the system terminal compensates the association analysis results based on the weakened association factor. The compensation process is to multiply the weakened association factor with the confidence, so that the influence of historical conversations with a long time is reduced, and the influence of historical conversations with a more recent time on the analysis results is greater, thereby improving the accuracy of the analysis results; after completing the compensation, the system terminal will create a context window with the compensated association analysis results, and the compensated confidence in the association analysis results will be used as the context trust mark of the context window to ensure that the AIGC robot can more accurately understand the background, emotion and intention of the current conversation, making the conversation more coherent and accurate. ;

[0032] Furthermore, the calculation of the reply decision value of the current dialogue round in the context window to generate the reply decision value calculation result includes:

[0033] ;in, Characterize the response decision value, is the probability of misunderstanding identified in the current dialogue, Represents the intention matching degree, Represents the emotional feedback score, , Characterizing time steps The dialogue text, Representing Dialogue Text The number of sub-fragments, For dialogue text Middle The sentiment score of each sub-segment, Characterizes the weighted value of historical behavior, , Characterize the user Behavioral characteristics in time The value of is the total number of behavioral characteristics, Characterization The weight coefficient of each behavior feature, Characterize the user interaction feature score, is a constant term, , , , , They are the weight factors of misunderstanding probability, intention matching, emotional feedback, historical behavior and interaction characteristics.

[0034] Optionally, when calculating the reply decision value of the current round of dialogue, the system terminal uses the reply decision value calculation formula. This reply decision value calculation formula comprehensively considers multiple factors of the current dialogue, including misunderstanding probability, intent matching, emotional feedback score, historical behavior weighted value and user interaction feature score, and obtains the final result in a linear weighted manner. The reply decision value calculation formula is as follows: ;in, The response decision value represents the comprehensive decision score calculated by the AIGC robot in the current conversation. This value determines the quality of the response given by the AIGC robot in the end. is the probability of misunderstanding identified by the AIGC robot in the current conversation. If P is higher, it means that the AIGC robot is more likely to misunderstand the current conversation; Indicates the degree of understanding of the user's intention by the AIGC robot. The closer the value is to 1, the higher the matching degree is. The feedback score that represents the user's emotion in the current conversation. The value range is [-1,1]. Negative values ​​indicate negative emotions, and positive values ​​indicate positive emotions. The absolute value of the emotion score is taken during calculation to ensure that the AIGC robot focuses on the intensity of the emotion rather than the direction of the emotion (positive or negative). The calculation method is: , Characterizing time steps The dialogue text, Representing Dialogue Text The number of sub-fragments, For dialogue text Middle The sentiment score of each sub-segment, For dialogue text The sub-fragments; The weighted value of historical behavior is used to represent the impact of the user's historical behavior on the current conversation. The calculation method is: , Characterize the user Behavioral characteristics in time The value of is the total number of behavioral characteristics, Characterization The weight coefficient of each behavioral feature; Feature scores that characterize user interaction behaviors, such as clicks, dwell time, and other data; is a constant term, which represents a fixed bias in the response decision calculation and is usually used to calibrate the final decision value; , , , , They are the weight factors of misunderstanding probability, intention matching, emotional feedback, historical behavior and interactive characteristics, indicating the importance of each factor in calculating the reply decision value. The size of the weight needs to be optimized according to the actual application scenario.

[0035] Furthermore, the establishment of the decision-making argumentation result of the response includes:

[0036] If the decision authentication result is an active confirmation result, a confirmation question set is established based on the dialogue, and the confirmation question set has a trust mark; the confirmation question set is actively confirmed one by one from high to low using the trust in the trust mark; when the active confirmation result meets the preset decision authentication threshold at any time, the decision authentication result is updated to the output result decision.

[0037] Preferably, after the decision authentication is completed by the decision layer of the AIGC robot, if the decision authentication result is judged to be an active confirmation result that requires further confirmation, the system terminal will generate a set of confirmation questions based on the current conversation content, context, and frequent item sets. The purpose of these questions is to verify whether the understanding of the current conversation is accurate and to ensure that the AIGC robot can provide a correct response. For example, if the creator's needs are unclear, the AIGC robot may ask some clarifying questions, such as whether to inquire about the current status of the fleet. Each question will be accompanied by a trust mark, indicating the degree of confidence of the AIGC robot in its understanding of the current conversation. The higher the trust, the more confident the AIGC robot is that the current understanding is correct. This trust mark is also based on the confidence. Confirmed; then, the system terminal will actively confirm the questions in the confirmation question set one by one from high to low according to the trust value in the trust identifier, and give priority to asking those questions with higher trust to the creator, because these questions have a higher degree of confidence, and answering these questions can more quickly verify whether the AIGC robot’s understanding of the current needs is correct; in the process of interacting with the creator, the AIGC robot will evaluate the creator’s feedback in real time by replying to the decision value calculation formula. When the result of the active confirmation meets the preset decision authentication threshold, the AIGC robot will stop confirming and believe that the current decision is accurate enough. At this time, the system terminal will update the decision authentication result from the active confirmation result state to the output result decision, thereby ensuring the accuracy and efficiency of the response.

[0038] When the replied decision authentication result is an output result decision, the multimodal data analysis network is called according to the output result decision to perform multidimensional business data integration and establish a multidimensional business data integration result.

[0039] In one embodiment, when the result of the judgment in the decision authentication process is an output result decision, the system terminal will enter the data analysis and integration stage. In this stage, the system terminal triggers the call of the multimodal data analysis network according to the output result decision. It is an analysis tool that can process various types of data, such as text data (user query instructions, historical conversation records, etc.), numerical data (business statistical data, fleet status data), historical records (user's past operating habits, query frequency), etc. The purpose is to integrate and comprehensively analyze information from different data sources. The multimodal data analysis network matches and integrates these multidimensional business data, associates and uniformly represents data of different dimensions through specific data analysis models (such as normalization, denoising, etc.), and generates a multidimensional business data integration result. The result is a high-quality data analysis result after multimodal data fusion, which provides support for subsequent output reports or decisions.

[0040] Furthermore, the decision to call the multimodal data analysis network to perform multidimensional business data integration according to the output result to establish the multidimensional business data integration result includes:

[0041] The multi-dimensional data matching is performed using the multi-modal data analysis network to establish multi-dimensional business data; the data analysis model in the multi-modal data analysis network is activated, and the multi-dimensional business data offline analysis is performed through the data analysis model to establish the multi-dimensional business data integration result, wherein the offline analysis includes data denoising, normalization processing, and joint authentication and identification.

[0042] Optionally, when the system terminal enters the multimodal data analysis stage, the data will be matched and analyzed through the multimodal data analysis network. First, the multimodal data analysis network is called to match and integrate multi-source and multi-type business data to establish a multi-dimensional business data set, such as user instructions, fleet operation status, statistical data, business indicators, etc., to facilitate subsequent analysis; after the multimodal data matching is completed, the system terminal will activate multiple data analysis models preset in the multimodal data analysis network, including a data processing model and a joint authentication and identification model, wherein the data processing model is responsible for basic data processing, such as denoising and normalization, and the joint authentication and identification model is used to verify the authenticity and consistency of the data to ensure data quality; by activating these data analysis models, the system terminal performs offline analysis of the multidimensional business data, and processes the text data and numerical data through the denoising algorithm in the data processing model, for example, removing punctuation marks, special characters, HTML tags, repeated content, etc. from the text data, and detecting and verifying outliers in the numerical data through the box plot method. Processing, use the mean, median or interpolation method to fill in missing values, and then normalize the denoised multidimensional business data through a normalization algorithm (such as the maximum and minimum method), and uniformly process the data in different data sources to make the data comparable; after the processing is completed, the processed multidimensional business data is verified for consistency and cross-checked through joint authentication and identification to ensure the authenticity and accuracy of the data. Taking hash verification as an example, a unique hash value is generated for the processed multidimensional business data using a hash algorithm to quickly determine whether the data content is consistent. If the hash values ​​are different, cross-checking is performed through feature matching (such as text semantic similarity, numerical error calculation). When the feature similarity reaches the set threshold, the data is deemed consistent, otherwise it is marked as abnormal; finally, the output marked as abnormal is removed from the processed multidimensional business data, thereby forming a multidimensional business data integration result. This result is a high-quality, verified data set that can reflect the relationship and key features between various business dimensions, providing reliable data support for subsequent report generation or decision-making.

[0043] After optimizing the multi-dimensional business data integration result according to the demand characteristics, an output report is generated, and creator feedback is obtained, and the adaptive feedback layer is updated using the creator feedback.

[0044] In one embodiment, after the multi-dimensional business data integration results are generated, the system terminal will perform optimization processes such as screening, sorting and processing on the integrated data according to the creator's demand characteristics (such as the focus of the report, the form of data display, etc.), that is, according to the demand characteristics, the corresponding data is extracted from the multi-dimensional business data integration results for prominent display. For example, when the creator's demand characteristics focus on the failure rate analysis of the fleet, the system terminal will extract key data related to the vehicle status from the multi-dimensional business data integration results, such as the vehicle's abnormal status, online status, fault conditions and other operating conditions; then, according to the creator's preference for the report display form, a corresponding visual report is generated. For example, when the demand is to display the distribution of data in a pie chart, the system terminal will configure the corresponding pie chart parameters, such as transparency, style, etc., and generate an output report in the form of a pie chart. report; when the demand is to analyze the proportion of fault alarm types, a bar chart is generated; if the creator is more concerned about the changing trend of the fleet status, a trend forecast chart will be generated to show the law of data changes over time; in this way, it can be ensured that the report content and display format can accurately meet the actual needs of the creator; after generating the output report, the system terminal will provide the creator with feedback channels (such as ratings, comments, selections, etc.) to collect the creator's satisfaction with the report content and improvement suggestions as explicit feedback. In addition, the creator's implicit feedback (such as viewing time, click operations, modification input, etc.) will be monitored. Based on the collected explicit feedback and implicit feedback, the system terminal will update the adaptive feedback layer, that is, perform incremental learning on the adaptive feedback layer, optimize the subsequent data integration and report generation process, and enable the AIGC robot to adapt to the needs of the creator more accurately.

[0045] Further, the updating the adaptive feedback layer using the creator feedback includes:

[0046] The creator feedback is classified into explicit feedback and implicit feedback, wherein the explicit feedback includes ratings, likes, comments, and selections, and the implicit feedback includes dwell time, click-through rate, changes in input text content, and emotional feedback; the explicit feedback and the implicit feedback are used to perform incremental learning of operation features and demand features, and the adaptive feedback layer is updated with the incremental learning results.

[0047] Preferably, the system terminal classifies the creator's feedback into two categories: explicit feedback and implicit feedback. Explicit feedback is information actively provided by the creator, including ratings, likes, comments, and selections that clearly express satisfaction or needs; implicit feedback is information collected indirectly through the creator's behavioral characteristics, such as the creator's residence time on the page, click-through rate, changes in input text content, and emotional feedback analyzed from the conversation; by collecting these explicit and implicit feedback, the system terminal uses forward propagation, loss calculation, back propagation, parameter optimization and other steps to perform incremental learning on the adaptive feedback layer, and gradually updates and optimizes the creator's operating characteristics and demand characteristics. For example, high-frequency likes or specific types of selections in explicit feedback can help the system terminal identify characteristics that are more in line with the creator's needs, while longer residence time or high click-through rate in implicit feedback can reflect the creator's high attention to certain information and reflect the creator's operating characteristics. The system terminal continuously improves the creator's adaptive feedback layer through incremental learning of these characteristics, so that the creator's needs can be more accurately met in subsequent report generation and interaction.

[0048] Furthermore, the output report can be called and distributed on multiple terminals through an open API interface.

[0049] Preferably, the output reports generated by the system terminal support calling and distribution through an open API interface, so that the output reports can be accessed and used on multiple ports. Whether it is the Web, mobile or other system integration platforms, the creator can access the report content through standardized API requests and obtain outputs in various forms including charts, data analysis results and business reports; the API interface provides flexible calling methods, such as obtaining specific types of reports on demand or pushing the latest data in real time, ensuring that the distribution of output reports is efficient and adaptable to a variety of application scenarios, realizing data sharing and cross-platform collaboration.

[0050] In summary, the embodiments of the present application have at least the following technical effects:

[0051] First, after creating the AIGC robot, the role logic and task objectives of the AIGC robot are defined according to the needs of the creator, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and the permission information. Subsequently, the initialized AIGC robot is used to conduct a dialogue with the creator, and an adaptive feedback layer is established, which is used to record the operating characteristics and demand characteristics of the creator. Further, the decision layer of the AIGC robot is called to perform the reply decision authentication of the dialogue, and the decision authentication result of the reply is established. Afterwards, when the decision authentication result of the reply is an output result decision, the multimodal data analysis network is called according to the output result decision to perform multidimensional business data integration, and a multidimensional business data integration result is established. Finally, after optimizing the multidimensional business data integration result according to the demand characteristics, an output report is generated, and the creator feedback is obtained, and the adaptive feedback layer is updated using the creator feedback. It solves the technical problems of insufficient context understanding depth, real-time performance and accuracy of generated content in existing artificial intelligence platforms based on large language models. By introducing identity and authority information reading, adaptive feedback layer and multimodal data analysis network, it establishes task-specific role logic and multi-dimensional business data integration mechanism, achieving the technical effect of enhancing the depth of context understanding, improving data generation accuracy and optimizing user interaction experience.

[0052] Embodiment 2 is based on the same inventive concept as the platform reporting method based on the AIGC robot in the above embodiment. Figure 2 As shown, the present application provides a platform reporting system based on an AIGC robot, wherein the system includes:

[0053] Robot initialization processing component 11: After creating the AIGC robot, define the role logic and task objectives of the AIGC robot according to the creator's needs, read the identity information and permission information of the creator, and use the identity information and permission information to initialize the AIGC robot; adaptive feedback layer establishment component 12: use the initialized AIGC robot to have a dialogue with the creator, and establish an adaptive feedback layer, the adaptive feedback layer is used to record the creator's operation characteristics and demand characteristics; reply decision authentication component 13: call the decision layer of the AIGC robot to perform reply decision authentication of the dialogue, and establish a reply decision authentication result; multi-dimensional business data integration component 14: when the reply decision authentication result is an output result decision, call the multimodal data analysis network according to the output result decision to perform multi-dimensional business data integration, and establish a multi-dimensional business data integration result; output report generation component 15: after optimizing the multi-dimensional business data integration result according to the demand characteristics, generate an output report, obtain creator feedback, and use the creator feedback to update the adaptive feedback layer.

[0054] Furthermore, the reply decision authentication component 13 is used to execute the following method:

[0055] The decision layer is used to perform dialogue backtracking and establish a backtracking data set; the dialogue features of the current dialogue are obtained, the dialogue features are used to perform time series correlation analysis of the backtracking data set, and the time series correlation results are used to create a context window, wherein the context window has a context trust mark; the reply decision value of the current dialogue is calculated within the context window to generate a reply decision value calculation result; after compensating the reply decision value calculation result through the context trust mark, the reply decision authentication is completed.

[0056] Furthermore, the reply decision authentication component 13 is used to execute the following method:

[0057] The data timestamp of the retrospective data set is obtained, and a weakened correlation factor is constructed using the data timestamp; an association rule algorithm is used to perform an association analysis between the retrospective data set and the user emotion, user behavior and user intention of the conversation feature, and an association analysis result is established; and the association analysis result is compensated based on the weakened correlation factor to complete the time series association analysis.

[0058] Furthermore, the reply decision authentication component 13 is used to execute the following method:

[0059] ;in, Characterize the response decision value, is the probability of misunderstanding identified in the current dialogue, Represents the intention matching degree, Represents the emotional feedback score, , Characterizing time steps The dialogue text, Representing Dialogue Text The number of sub-fragments, For dialogue text Middle The sentiment score of each sub-segment, Characterizes the weighted value of historical behavior, , Characterize the user Behavioral characteristics in time The value of is the total number of behavioral characteristics, Characterization The weight coefficient of each behavior feature, Characterize the user interaction feature score, is a constant term, , , , , They are the weight factors of misunderstanding probability, intention matching, emotional feedback, historical behavior and interaction characteristics.

[0060] Furthermore, the reply decision authentication component 13 is used to execute the following method:

[0061] If the decision authentication result is an active confirmation result, a confirmation question set is established based on the dialogue, and the confirmation question set has a trust mark; the confirmation question set is actively confirmed one by one from high to low using the trust in the trust mark; when the active confirmation result meets the preset decision authentication threshold at any time, the decision authentication result is updated to the output result decision.

[0062] Furthermore, the multi-dimensional business data integration component 14 is used to execute the following method:

[0063] The multi-dimensional data matching is performed using the multi-modal data analysis network to establish multi-dimensional business data; the data analysis model in the multi-modal data analysis network is activated, and the multi-dimensional business data offline analysis is performed through the data analysis model to establish the multi-dimensional business data integration result, wherein the offline analysis includes data denoising, normalization processing, and joint authentication and identification.

[0064] Furthermore, the output report generating component 15 is used to execute the following method:

[0065] The creator feedback is classified into explicit feedback and implicit feedback, wherein the explicit feedback includes ratings, likes, comments, and selections, and the implicit feedback includes dwell time, click-through rate, changes in input text content, and emotional feedback; the explicit feedback and the implicit feedback are used to perform incremental learning of operation features and demand features, and the adaptive feedback layer is updated with the incremental learning results.

[0066] Furthermore, the output report generating component 15 is used to execute the following method:

[0067] The output report can be called and distributed on multiple terminals through an open API interface.

[0068] Embodiment 3 provides a storage medium, wherein a computer program is stored in the medium, and when the processor executes the computer program, the following method is implemented:

[0069] After the AIGC robot is created, the role logic and task objectives of the AIGC robot are defined according to the needs of the creator, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and the permission information; the initialized AIGC robot is used to conduct a dialogue with the creator, and an adaptive feedback layer is established, and the adaptive feedback layer is used to record the operating characteristics and demand characteristics of the creator; the decision layer of the AIGC robot is called to authenticate the reply decision of the dialogue, and the decision authentication result of the reply is established; when the decision authentication result of the reply is an output result decision, a multimodal data analysis network is called according to the output result decision to perform multidimensional business data integration, and a multidimensional business data integration result is established; after optimizing the multidimensional business data integration result according to the demand characteristics, an output report is generated, and feedback from the creator is obtained, and the adaptive feedback layer is updated using the creator feedback.

[0070] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0072] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A platform reporting method based on AIGC robot, characterized in that: The method comprises: After the AIGC robot is created, the role logic and task objectives of the AIGC robot are defined according to the creator's needs, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and permission information; The initialized AIGC robot is used to conduct a dialogue with the creator, and an adaptive feedback layer is established, wherein the adaptive feedback layer is used to record the operating characteristics and demand characteristics of the creator; Calling the decision layer of the AIGC robot to perform reply decision authentication of the dialogue, and establishing a reply decision authentication result; When the replied decision authentication result is an output result decision, a multimodal data analysis network is called according to the output result decision to perform multidimensional business data integration and establish a multidimensional business data integration result; After optimizing the multi-dimensional business data integration result according to the demand characteristics, an output report is generated, and creator feedback is obtained, and the adaptive feedback layer is updated using the creator feedback; The calling of the decision layer of the AIGC robot to perform a reply decision authentication of the dialogue and establish a reply decision authentication result includes: Using the decision layer to conduct dialogue backtracking and establish a backtracking data set; Acquire the dialogue features of the current dialogue, use the dialogue features to perform time series correlation analysis of the backtracking data set, and use the time series correlation results to create a context window, wherein the context window has a context trust mark; Calculating the reply decision value of the current dialogue round in the context window to generate a reply decision value calculation result; After compensating the reply decision value calculation result through the context trust identifier, the reply decision authentication is completed.

2. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: The using the conversation features to perform temporal correlation analysis of the retrospective data set includes: Obtain the data timestamp of the retrospective data set and use the data timestamp to construct a weakened correlation factor; Using an association rule algorithm to perform an association analysis between the backtracking data set and the user emotions, user behaviors, and user intentions of the conversation features, and establish an association analysis result; The correlation analysis result is compensated based on the weakened correlation factor to complete the time series correlation analysis.

3. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: The step of calculating the reply decision value of the current dialogue round in the context window and generating a reply decision value calculation result includes: ； in, Characterize the response decision value, is the probability of misunderstanding identified in the current dialogue, Represents the intention matching degree, Represents the emotional feedback score, , Characterizing time steps The dialogue text, Representing Dialogue Text The number of sub-fragments, For dialogue text Middle The sentiment score of each sub-segment, Characterizes the weighted value of historical behavior, , Characterize the user The behavior features are in the time step The value of is the total number of behavioral characteristics, Characterization The weight coefficient of each behavior feature, Characterize the user interaction feature score, is a constant term, , , , , They are the weight factors of misunderstanding probability, intention matching, emotional feedback, historical behavior and interaction characteristics.

4. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: After the decision-making and argumentation results of the response are established, it includes: If the decision authentication result is an active confirmation result, a confirmation question set is established according to the dialogue, and the confirmation question set has a trust level mark; Actively confirm the set of confirmation questions one by one from high to low using the trust levels in the trust level identifier; When the active confirmation result at any time meets the preset decision authentication threshold, the decision authentication result is updated to output the result decision.

5. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: The decision to call the multimodal data analysis network to perform multidimensional business data integration according to the output result to establish the multidimensional business data integration result includes: Using the multimodal data analysis network to perform multidimensional data matching and establish multidimensional business data; Activate the data analysis model in the multimodal data analysis network, perform offline analysis of multidimensional business data through the data analysis model, and establish the integration result of the multidimensional business data. The offline analysis includes data denoising, normalization processing, and joint authentication and identification.

6. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: The updating of the adaptive feedback layer using the creator feedback comprises: Classify the creator feedback to establish explicit feedback and implicit feedback, where the explicit feedback includes ratings, likes, comments, and selections, and the implicit feedback includes dwell time, click-through rate, changes in input text content, and emotional feedback; The explicit feedback and the implicit feedback are used to perform incremental learning of operation characteristics and demand characteristics, and the adaptive feedback layer is updated with the incremental learning results.

7. A platform reporting method based on an AIGC robot as claimed in claim 1, characterized in that: The output report can be called and distributed on multiple terminals through an open API interface.

8. A platform reporting system based on AIGC robot, characterized in that: A platform reporting method based on an AIGC robot for implementing any one of claims 1 to 7, the system comprising: Robot initialization processing component: After the AIGC robot is created, the role logic and task objectives of the AIGC robot are defined according to the creator's needs, and the identity information and permission information of the creator are read, and the AIGC robot is initialized with the identity information and permission information; Adaptive feedback layer establishment component: using the initialized AIGC robot to conduct a dialogue with the creator and establish an adaptive feedback layer, the adaptive feedback layer is used to record the operating characteristics and demand characteristics of the creator; Reply decision authentication component: calls the decision layer of the AIGC robot to perform reply decision authentication for the conversation and establishes a reply decision authentication result; Multi-dimensional business data integration component: when the decision authentication result replied is an output result decision, the multi-modal data analysis network is called according to the output result decision to perform multi-dimensional business data integration and establish a multi-dimensional business data integration result; Output report generation component: after optimizing the multi-dimensional business data integration result according to the demand characteristics, generate an output report, obtain creator feedback, and use the creator feedback to update the adaptive feedback layer.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements a platform reporting method based on an AIGC robot as described in any one of claims 1 to 7.

Images