Time limit early warning monitoring and supervising method and system based on AI technology

Abstract

The application relates to the technical field of time limit early warning management and control, and particularly discloses a time limit early warning monitoring and supervising method and system based on AI technology, which first extracts a target task from a task library, and analyzes the target task by using a natural language processing technology to extract the task type, the deadline and the information of the person in charge of the target task, and further judges the urgency and importance of the task, and then sets the early warning level and frequency of the target task according to the urgency and importance of the task, and when the distance to the deadline of the task reaches the set early warning level, the monitoring and supervising mechanism is automatically triggered to send early warning information to the relevant person in charge at the set frequency to check and remind the task progress. In this way, the efficiency and effect of the supply and service task management can be effectively improved, and the loss caused by the task delay can be reduced.

Description

Technical Field

[0001] This application relates to the field of time-limited early warning and control technology, and more specifically, to a time-limited early warning monitoring and supervision method and system based on AI technology. Background Technology

[0002] With the development of the social economy and the increasing demands of users for power service quality, power supply companies face more complex and ever-changing working environments and service challenges. In the current power supply service command system, task management not only requires high efficiency and accuracy, but also needs to ensure that every task is completed on time and with high quality to guarantee the continuity and stability of power supply services.

[0003] However, traditionally, the timeliness and control of power supply tasks have relied primarily on manual operation. Power supply service personnel need to search through multiple different menus and systems to collect tasks that are about to expire or have already exceeded their deadlines, and then use communication tools such as telephone and WeChat to convey the supervision information to the relevant responsible personnel one by one. This process is not only time-consuming, labor-intensive, and inefficient, making it difficult to guarantee the accuracy and timeliness of information, but it is also prone to task omissions or delays due to human negligence, further increasing the risk of task delays.

[0004] In recent years, information technology has been widely applied in the power supply service sector, but its level of intelligence in time-limit early warning and management remains low. Specifically, existing task management systems mainly rely on fixed time reminder settings for task time-limit early warning, lacking intelligent early warning mechanisms. They cannot automatically adjust supervision strategies based on the urgency and importance of tasks, which may lead to untimely responses to urgent tasks.

[0005] Therefore, we look forward to a time-limited early warning monitoring and supervision method and system based on AI technology. Summary of the Invention

[0006] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a time-limit warning monitoring and supervision method and system based on AI technology. First, target tasks are extracted from a task database, and natural language processing technology is used to parse the target tasks to extract their task type, deadline, and responsible person information. The urgency and importance of the task are further determined. Then, based on the urgency and importance, a warning level and frequency are set for the target task. When the time remaining until the task deadline reaches the set warning level, a monitoring and supervision mechanism is automatically triggered, sending warning information to relevant responsible persons at a set frequency to check and remind them of the task progress. In this way, the efficiency and effectiveness of supply chain task management can be effectively improved, reducing losses caused by task delays.

[0007] According to one aspect of this application, a time-limited early warning monitoring and supervision method based on AI technology is provided, which includes:

[0008] Retrieve the first target task from the task library;

[0009] Natural language processing techniques are used to parse the first target task to obtain the text parsing result of the first target task;

[0010] Extract key information of the first target task from the text parsing result of the first target task, including task type, deadline, and person in charge;

[0011] Based on the text parsing results of the first target task, determine the urgency and importance tags of the first target task;

[0012] Based on the urgency and importance tags of the first target task, set the warning level and frequency;

[0013] Based on the comparison between the current date and the deadline, and based on the warning level and frequency, the first target task is monitored and supervised.

[0014] According to another aspect of this application, a time-limit early warning monitoring and supervision system based on AI technology is provided, comprising:

[0015] The first target task extraction module is used to extract the first target task from the task library;

[0016] The text parsing module is used to parse the first target task using natural language processing technology to obtain the text parsing result of the first target task;

[0017] The key information extraction module is used to extract key information of the first target task from the text parsing result of the first target task. The key information includes task type, deadline, and person in charge.

[0018] The priority assessment module is used to determine the urgency and importance tags of the first target task based on the text parsing results of the first target task.

[0019] The warning level setting module is used to set the warning level and frequency based on the urgency and importance tags of the first target task;

[0020] The task monitoring and supervision module is used to monitor and supervise the first target task based on the comparison between the current date and the deadline, and based on the warning level and frequency.

[0021] Compared with existing technologies, the time-limit warning monitoring and supervision method and system based on AI technology provided in this application first extracts target tasks from a task database and uses natural language processing technology to parse the target tasks to extract the task type, deadline, and responsible person information. It then further determines the urgency and importance of the task. Based on the urgency and importance, it sets the warning level and frequency for the target task. When the time remaining until the task deadline reaches the set warning level, the monitoring and supervision mechanism is automatically triggered, sending warning information to relevant responsible persons at the set frequency to check and remind them of the task progress. This approach can effectively improve the efficiency and effectiveness of supply chain task management and reduce losses caused by task delays. Attached Figure Description

[0022] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 This is a flowchart of a time-limited early warning monitoring and supervision method based on AI technology according to an embodiment of this application.

[0024] Figure 2 This is a schematic diagram of data flow in the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application.

[0025] Figure 3 This is a flowchart of sub-step S4 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application.

[0026] Figure 4 This is a flowchart of sub-step S43 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application.

[0027] Figure 5 This is a flowchart of sub-step S431 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application.

[0028] Figure 6 This is a block diagram of an AI-based time-limited early warning monitoring and supervision system according to an embodiment of this application. Detailed Implementation

[0029] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0030] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0031] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0032] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0033] It is worth noting that all data acquisition actions in this application were carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization granted by the owner of the relevant device.

[0034] To address the technical problems described in the background, this application proposes an AI-based method for time-limit early warning monitoring and supervision. First, target tasks are extracted from a task database, and natural language processing (NLP) is used to parse these tasks, extracting their task type, deadline, and responsible person information. The urgency and importance of the task are then assessed. Based on these urgency and importance, an early warning level and frequency are set for the target task. When the deadline reaches the set early warning level, a monitoring and supervision mechanism is automatically triggered, sending early warning notifications to relevant responsible persons at a set frequency to check and remind them of task progress. This approach effectively improves the efficiency and effectiveness of supply chain task management and reduces losses caused by task delays.

[0035] Figure 1 This is a flowchart of a time-limited early warning monitoring and supervision method based on AI technology according to an embodiment of this application. Figure 2This is a schematic diagram of the data flow in the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application. Figure 1 and Figure 2 As shown, the AI-based time-limited early warning monitoring and supervision method includes the following steps: S1, extracting a first target task from a task library; S2, parsing the first target task using natural language processing technology to obtain a text parsing result of the first target task; S3, extracting key information of the first target task from the text parsing result of the first target task, the key information including task type, deadline, and person in charge; S4, determining the urgency label and importance label of the first target task based on the text parsing result of the first target task; S5, setting an early warning level and frequency based on the urgency label and importance label of the first target task; S6, monitoring and supervising the first target task based on a comparison between the current date and the deadline, and based on the early warning level and frequency.

[0036] In the aforementioned AI-based time-limited early warning monitoring and supervision method, step S1 involves retrieving the first target task from the task library. It should be understood that the task library is a collection of all tasks within the power supply company, storing detailed information about various tasks, including but not limited to task name, task type, task status, task description, deadline, and responsible person.

[0037] Specifically, in the management of power supply service tasks, establishing a structured and easily accessible task database is crucial. By rationally designing the task database structure and creating indexes for frequently used query fields (such as task name, due date, and responsible person), query efficiency can be significantly improved. This design not only facilitates rapid information retrieval but also ensures data consistency and integrity.

[0038] Next, ensuring system security and reliability is essential. This involves strict access control, using Role-Based Access Control (RBAC) to manage access permissions for different users to guarantee the security of sensitive information. Different user roles are assigned different operational permissions, ensuring that only authorized personnel can perform specific operations. Furthermore, regular data backups and detailed data recovery plans are crucial to prevent data loss or corruption. Backup strategies should include both full and incremental backups to reduce storage space usage and shorten recovery time.

[0039] After preparing the task library, the process of extracting the "first target task" can be achieved through a simple SQL query. The goal here is to select the most urgent and important tasks from the task library. Assuming "first" is defined as one or more tasks with the highest priority determined by business logic, then a list of tasks that meet specific conditions can be directly retrieved from the database by writing an SQL query.

[0040] Furthermore, if the task management system provides an API, these APIs can be called using programming languages ​​(such as Python, JavaScript, etc.) to obtain the required data. For example, query parameters can be sent via HTTP requests, and query conditions can be dynamically adjusted to adapt to different business needs. This approach is not only flexible but also allows for rapid integration into existing workflows. APIs can also be used to automate task assignment, automatically notifying relevant personnel when a new task is added to the system, reducing manual intervention.

[0041] For task scheduling scenarios that don't require immediate response, cron jobs can be set up to periodically execute batch processing programs. These programs can run automatically at preset time intervals, checking all pending items in the task library and evaluating and selecting the most critical tasks according to established rules. This method is particularly suitable for long-term monitoring tasks running in the background, helping to reduce the burden of manual operation. Cron jobs can be implemented using the operating system's built-in task scheduling tools (such as Linux's cron) or application-level schedulers (such as Quartz).

[0042] Finally, considering scenarios with high real-time requirements, an event-driven architecture can be considered. When a new task is added or the state of an existing task changes, a corresponding event notification is triggered. Application components listening for these events can automatically filter task instances that match the definition of the "first target task" based on preset rules. This can be implemented using message queues or similar technologies, providing a lightweight approach to handling real-time changes. Message queues such as RabbitMQ or Kafka can ensure reliable delivery of event notifications, maintaining system stability even in the event of network failures or other anomalies.

[0043] In summary, by constructing a structured task library, utilizing SQL queries for efficient retrieval, combining API calls, and setting up scheduled tasks and an event-driven architecture, it is possible to accurately extract the "primary target task" from the task library. This effectively improves work efficiency and lays a solid foundation for subsequent data analysis and processing.

[0044] In the above time limit warning monitoring and supervision method based on AI technology, in step S2, natural language processing technology is used to parse the first target task to obtain the parsing result of the first target task text. It should be understood that considering that the original task information in the task library may follow different templates or format requirements, in order to achieve unified task information processing and analysis, this application uses natural language processing technology to parse the first target task to convert it into a standardized task parsing result in text form, which is convenient for subsequent data processing. In the embodiment of this application, a GPT series model is used to parse the first target task to generate the parsing result of the first target task text.

[0045] Using natural language processing (NLP) technology to parse the first target task is a key step in realizing intelligent time limit warning control. This process starts with data preprocessing. Through a series of text cleaning and formatting operations, a clean and unified foundation is provided for subsequent natural language processing algorithms. At this stage, irrelevant characters or symbols in the original task text will be removed, such as HTML tags, extra spaces or punctuation marks, and different forms of expressions will be unified, such as date formats and capitalization, to ensure that all dates adopt a unified format and the consistency of capitalization is also guaranteed. At the same time, the continuous text is segmented into meaningful lexical units through word segmentation. For Chinese task texts, word segmentation is particularly important because Chinese has no clear word boundaries. In addition, high-frequency words that do not carry actual meaning (i.e., stop words), such as "de", "shi", "zai", etc., need to be removed to reduce interference; and the words are restored to their basic forms, for example, verb conjugations are restored to the original form, so that different forms represent the same word.

[0046] First, in order to ensure that the GPT model can accurately understand and parse the task description, it is necessary to perform appropriate preparation and preprocessing on the input data. This step includes formatting the input text, converting the first target task description extracted from the task library into a form suitable for the GPT model to process, such as removing extra spaces, standardizing punctuation marks, etc. At the same time, considering the possible structured features of the task description, constructing one or more prompt templates to standardize the input form helps to improve the parsing accuracy. In addition, for the terms and technical vocabulary unique to the power industry, the vocabulary of the GPT model is pre-expanded to enable it to better understand the expressions in a specific field.

[0047] Based on this prepared data, select a suitable GPT model version (such as GPT-3 or a later version) and optimize it according to the specific application scenario. The optimization process involves fine-tuning the model using a large amount of historical task description data from the power industry to enhance its understanding of domain-specific languages ​​and optimize the accuracy of identifying key task information. Simultaneously, appropriately set the model's context window size to ensure complete capture of relevant information for each task description, and if multilingual task descriptions exist, ensure the selected model supports multilingual parsing.

[0048] After the above preparations, we can begin using the GPT model to analyze the first target task. This process automatically identifies the main intent in the task description through the powerful semantic understanding capabilities of the GPT model. Then, using Named Entity Recognition (NER) technology combined with the contextual understanding capabilities of the GPT model, we accurately extract key entities from the task description, such as time, location, and personnel names. This is crucial for subsequently determining the key information of the task. Next, we leverage the logical reasoning capabilities of the GPT model to infer the potential relationships between the elements in the task description, such as determining whether a certain point in time is the start or end time of the task.

[0049] Finally, the output format of the parsed results obtained from the GPT model is defined. Considering the needs of practical applications, the output should be a structured text data object containing all the parsed key information. This structured output can be directly used for the next step of extracting key task information and other related processing. The output format should be designed to be concise and easy to integrate into existing task management systems.

[0050] By using the GPT series models in this way to analyze the first target task, not only can the efficiency and accuracy of the analysis be significantly improved, but an important foundation can also be laid for realizing intelligent time-limited early warning and control.

[0051] In the aforementioned AI-based time-limit warning, monitoring, and supervision method, step S3 involves extracting key information about the first target task from the text parsing results. This key information includes the task type, deadline, and responsible person. It should be understood that the task type defines the nature and execution standards of the task; understanding the task type of the first target task facilitates task classification and management. The deadline is a crucial time constraint for the task, directly related to its urgency and priority. Identifying the deadline allows for timely warnings and supervision. The responsible person is the key figure in task execution; clearly identifying the responsible person helps ensure the accurate communication of task warning and supervision information.

[0052] After using the GPT model to parse the first target task and obtain the text parsing results, the next step is to extract key information from the parsing results. This key information includes the task type, deadline, and person in charge, which is crucial for subsequent task management and early warning control.

[0053] First, after obtaining the text parsing results for the primary objective task, the unstructured text is converted into a structured data format. This step can be achieved by tagging each parsed information element, for example, by organizing this information using XML or JSON format, where each tag represents a specific category, such as...<task_type> (Task type) <deadline>(expiration date),<responsible_person> (Responsible Person), etc. This structure not only facilitates computer processing but also enables subsequent data analysis and application integration. To improve the accuracy of information extraction, it is necessary to develop or select appropriate key information identification algorithms. These algorithms can understand complex expressions in natural language and extract precise information from them. For example, the identification of deadlines may involve date pattern matching and contextual understanding to ensure that time information is correctly captured even in complex sentence structures; while the identification of responsible persons relies on Named Entity Recognition (NER) technology, combined with a list of personnel names, to improve identification accuracy.

[0054] In practice, some anomalies are inevitable, such as ambiguous task descriptions or missing key information. Therefore, establishing an effective anomaly detection system is crucial. This system should automatically identify potential problems and trigger corresponding warnings. Simultaneously, a user feedback mechanism should be designed to allow manual intervention to correct erroneous parsing results, ensuring the accuracy of the final extracted key information. This plays an indispensable role in maintaining system reliability and user trust. Finally, to guarantee the quality of the entire extraction process, a comprehensive testing and evaluation system must be built. This system should include multiple levels of checkpoints, covering everything from the accuracy of individual fields to the consistency of overall information. Model performance should be evaluated regularly, comparing improvements between different versions and adjusting optimization strategies accordingly.

[0055] In summary, extracting key information from the parsed text of the first target task is a process that comprehensively utilizes multiple technologies and methods. Through data structuring, feature annotation, selection and development of key information identification algorithms, application of logical reasoning, construction of anomaly detection and user feedback mechanisms, and rigorous testing and evaluation, the three key pieces of information—task type, deadline, and responsible person—can be effectively extracted from the parsing results, providing strong support for task management in the power supply service command system.

[0056] In the aforementioned AI-based time-limit warning monitoring and supervision method, step S4 determines the urgency and importance tags of the first target task based on the text parsing results. It should be understood that, considering the significant differences in complexity and workload among different tasks, relying solely on the task's deadline for time-limit warning monitoring and supervision may not be comprehensive or accurate enough, increasing the risk of task incompleteness. Therefore, this application further performs contextual semantic analysis on the text parsing results of the first target task to determine the urgency and importance of the first target task, thereby setting its warning level and frequency based on the task's urgency and importance, achieving more intelligent task management. Figure 3 This is a flowchart of sub-step S4 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application. Figure 3 As shown, step S4 includes the following steps: S41, performing word-granular semantic embedding encoding on the parsing result of the first target task text to obtain a sequence of word-granular semantic embedding encoding vectors for the first target task; S42, performing inter-word local semantic association encoding and global contextual semantic association encoding on the sequence of word-granular semantic embedding encoding vectors for the first target task to obtain an inter-word semantic association encoding vector and a contextual semantic encoding vector for the first target task; S43, performing fine-grained interactive fusion on the inter-word semantic association encoding vector and the contextual semantic encoding vector for the first target task to obtain a multi-scale semantic encoding vector for the first target task; S44, determining the urgency label and importance label of the first target task based on the multi-scale semantic encoding vector for the first target task.

[0057] Specifically, step S41 involves performing word-granular semantic embedding encoding on the first target task text parsing result to obtain a sequence of first target task word-granular semantic embedding encoding vectors. In a specific example of this application, step S41 includes: performing word segmentation on the first target task text parsing result to obtain a text distribution sequence of first target task word-granular descriptions; and inputting each first target task word-granular description in the text distribution sequence of the first target task word-granular descriptions into a BERT-based word embedding encoder to obtain a sequence of first target task word-granular semantic embedding encoding vectors. More specifically, to gain a more detailed understanding of the content meaning of the first target task text parsing result, this application further performs word segmentation on the first target task text parsing result to decompose continuous natural language text into independent lexical units, generating a text distribution sequence of first target task word-granular descriptions, thereby refining the granularity of semantic analysis and providing a foundation for subsequent semantic parsing. Next, in order to fully capture the deep semantic information of each lexical unit, this application uses a word embedding encoder based on the BERT model to perform word embedding encoders on each first target task word granular description. By utilizing the pre-trained knowledge of the BERT model, each lexical unit is transformed into a vector representation in a high-dimensional semantic space, thereby obtaining a sequence of first target task word granular semantic embedding encoding vectors with rich semantic information.

[0058] Specifically, step S42 involves performing inter-word local semantic association encoding and global contextual semantic association encoding on the sequence of the first target task word-granular semantic embedding encoding vector to obtain the first target task word-to-word semantic association encoding vector and the first target task contextual semantic encoding vector. In a specific example of this application, step S42 includes: inputting the sequence of the first target task word-granular semantic embedding encoding vector into an inter-word semantic information extraction network based on a one-dimensional convolutional layer to obtain the first target task word-to-word semantic association encoding vector. That is, this application considers that there are usually certain semantic associations and dependencies between various lexical units in the text parsing results of the first target task. Therefore, in order to more accurately understand the contextual meaning of the task, this application uses an inter-word semantic information extraction network based on a one-dimensional convolutional layer to process the sequence of the first target task word-granular semantic embedding encoding vector to obtain the first target task word-to-word semantic association encoding vector. It should be understood that by using a one-dimensional convolutional layer to perform a sliding window convolution operation on the sequence of word-granular semantic embedding encoding vectors for the first target task, the local semantic associations between lexical units can be effectively captured, and the semantic features of word combinations within a local range can be mined. This enhances the context awareness of the text parsing results of the first target task and provides more accurate semantic support for subsequent judgments on the urgency and importance of the task.

[0059] Specifically, in a specific example of this application, step S42 further includes: inputting the sequence of the first target task word-granular semantic embedding encoding vector into a converter-based global contextual semantic encoder to obtain the first target task contextual semantic encoding vector. It should be understood that, considering that convolution operations can only capture lexical semantic association information within a local range and ignore the semantic dependencies between long-distance words, in order to more comprehensively understand the context and background of the first target task text parsing results, this application further introduces a converter-based global contextual semantic encoder to perform global semantic association encoding on the sequence of the first target task word-granular semantic embedding encoding vector to obtain the first target task contextual semantic encoding vector. Those skilled in the art should know that converter models can capture the semantic dependencies between any two lexical units in the input sequence through a self-attention mechanism, regardless of how far apart they are in the text sequence. This global contextual encoding capability enables the model to more fully understand the global contextual information of the first target task, thereby providing more accurate information support for assessing the urgency and importance of the task.

[0060] Specifically, step S43 involves fine-grained interactive fusion of the first target task word semantic association encoding vector and the first target task context semantic encoding vector to obtain a first target task multi-scale semantic encoding vector. That is, to comprehensively consider the local semantic structure and global contextual information of the first target task text parsing result, this application further performs feature fusion on the first target task word semantic association encoding vector and the first target task context semantic encoding vector to provide a more comprehensive understanding of the task description. In particular, to improve the semantic fusion effect of multi-scale features, this application introduces external knowledge to guide the interactive fusion between the first target task word semantic association encoding vector and the first target task context semantic encoding vector, thereby further enhancing the model's ability to understand task features. Figure 4 This is a flowchart of sub-step S43 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application. Figure 4 As shown, step S43 includes the following steps: S431, performing feature modulation optimization based on external knowledge on the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector to obtain optimized first target task inter-word semantic association encoding vector and optimized first target task context semantic encoding vector; S432, performing position-by-position semantic interactive encoding on the optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector to obtain first target task multi-scale semantic encoding vector.

[0061] Figure 5 This is a flowchart of sub-step S431 of the AI-based time-limited early warning monitoring and supervision method according to an embodiment of this application. Figure 5 As shown, step S431 includes the following steps: S4311, inputting the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector into a fine-grained feature interaction network to obtain a first target task multi-scale semantic feature fine-grained interaction matrix; S4312, inputting the first target task multi-scale semantic feature fine-grained interaction matrix into an attention unit based on external knowledge to obtain an external knowledge-optimized first target task multi-scale semantic feature fine-grained interaction matrix; S4313, based on the external knowledge-optimized first target task multi-scale semantic feature fine-grained interaction matrix, performing feature modulation optimization on the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector respectively to obtain the optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector.

[0062] In a specific example of this application, step S4311 is expressed by the formula:

[0063]

[0064] Wherein, V1 represents the semantic association encoding vector between words in the first target task, V2 represents the semantic encoding vector of the context of the first target task, and (·) T Represents the transpose of a vector. M represents matrix multiplication. p This represents the fine-grained interaction matrix of multi-scale semantic features for the first objective task.

[0065] In other words, a fine-grained feature interaction network is used to perform a micro-level semantic association understanding on the semantic association encoding vector between words in the first target task and the semantic association encoding vector of the context of the first target task, so as to capture the fine-grained semantic interaction information between the two and generate a multi-scale semantic feature fine-grained interaction matrix of the first target task.

[0066] In a specific example of this application, step S4312 is expressed by the formula:

[0067]

[0068] Among them, M k and M v M represents the learnable memory parameter matrix of the attention unit based on external knowledge, norm(·) represents the normalization function, and M y This represents the fine-grained interaction matrix of multi-scale semantic features for the primary objective task of optimizing external knowledge.

[0069] In other words, the generated fine-grained interaction matrix of multi-scale semantic features for the first target task is fed into an attention unit based on external knowledge. This external knowledge is then used to further optimize the fine-grained interaction matrix, thereby enhancing the model's understanding of knowledge related to specific types of service tasks. In practice, an external knowledge base can be constructed for different types of service tasks, such as business rules, operational procedures, and service standards. Then, based on the task type of the first target task, relevant external knowledge is extracted from the knowledge base to train the parameter matrix of the attention unit. This ensures that the attention unit can accurately identify and strengthen semantic feature interaction patterns related to the urgency and importance of the task, guiding the attention unit to perform weighted optimization of the fine-grained interaction matrix of multi-scale semantic features for the first target task, thereby capturing key semantic information in the task text over a wider range.

[0070] In a specific example of this application, step S4313 includes: performing a linear transformation on the first target task inter-word semantic association encoding vector to obtain a first query feature vector and a first value feature vector, and using the external knowledge to optimize the first target task multi-scale semantic feature fine-grained interaction matrix as a key matrix; inputting the first query feature vector, the first value feature vector, and the key matrix into a Transformer-based fine-grained modulation module to obtain the optimized first target task inter-word semantic association encoding vector; performing a linear transformation on the first target task context semantic encoding vector to obtain a second query feature vector and a second value feature vector, and using the external knowledge to optimize the first target task multi-scale semantic feature fine-grained interaction matrix as a key matrix; inputting the second query feature vector, the second value feature vector, and the key matrix into the Transformer-based fine-grained modulation module to obtain the optimized first target task context semantic encoding vector, expressed by the formula:

[0071]

[0072] Among them, W 1q and W 1v V represents the first query embedding matrix and the first value embedding matrix, respectively. 1q and V 1v Let w represent the first query feature vector and the first value feature vector, respectively. 2q and w 2v V represents the second query embedding matrix and the second value embedding matrix, respectively. 2q and v 2v b represents the second query feature vector and the second value feature vector, respectively. 1q b 1v b 2q and b 2v These represent different bias terms, softmax(·) represents the normalization exponential function, d represents the feature scale value of the fine-grained feature interaction matrix optimized by the external knowledge, and V'1 and V'2 represent the optimized inter-word semantic association encoding vector and the optimized context semantic encoding vector of the first target task, respectively.

[0073] In other words, the fine-grained interaction matrix of the multi-scale semantic features of the first target task, optimized by external knowledge, is used as the key matrix. At the same time, a first query feature vector and a first value feature vector are constructed based on the semantic association encoding vector between words of the first target task. A second query feature vector and a second value feature vector are constructed based on the semantic encoding vector of the context of the first target task. The self-attention mechanism of the Transformer structure is used to realize the exchange and integration of information between the internal and external knowledge of the features, thereby ensuring that the semantic association encoding vector between words of the first target task and the semantic encoding vector of the context of the first target task can benefit from external knowledge to improve the quality of their feature representation.

[0074] More specifically, step S423 includes: calculating the position-point division between the optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector to obtain the first target task multi-scale semantic encoding vector, expressed by the formula:

[0075]

[0076] Among them, V i This represents the multi-scale semantic encoding vector of the first target task.

[0077] In other words, by dividing by position, the optimized semantic association encoding vector between words of the first target task and the optimized semantic context encoding vector of the first target task are semantically interactively encoded position by position to obtain the multi-scale semantic encoding vector of the first target task. This is used to comprehensively reflect the multi-level semantic information of the text parsing results of the first target task and to achieve a deep understanding of the first target task.

[0078] Specifically, step S44 involves determining the urgency label and importance label of the first target task based on the first target task multi-scale semantic encoding vector. In a specific example of this application, step S44 includes: inputting the first target task multi-scale semantic encoding vector into a classifier-based task recognizer to obtain the urgency label and importance label of the first target task. Specifically, the classifier employs a deep neural network structure. During training, a large amount of labeled task data is used as the training set to ensure that the model can learn the distinguishing features between different urgency and importance labels. The model parameters are continuously adjusted through backpropagation and gradient descent optimization methods to minimize the difference between the predicted label and the true label, thereby achieving an accurate assessment of the urgency and importance of the task. In particular, in the technical solution of this application, the classifier includes two output layers, corresponding to the prediction of the urgency label and the importance label, respectively. Each output layer uses a softmax activation function to ensure that the obtained probability distribution reflects the probability that the task belongs to different urgency and importance categories. In this way, it is possible to effectively determine the urgency and importance categories of tasks, such as high priority, medium priority, or low priority, as well as critical, important, or general importance.

[0079] In a preferred embodiment of this application, passing the multi-scale semantic encoding vector of the first target task through a classifier-based task recognizer to obtain the urgency label and importance label of the first target task includes:

[0080] First, the sum of the absolute values ​​of all feature values ​​of the first target task's multi-scale semantic coding vector is calculated to obtain the multi-scale semantic coding space structure value of the first target task. Then, the square root of the sum of squares is calculated to obtain the multi-scale semantic coding space structure value of the second target task, expressed by the formula:

[0081] w1=∑ i＝1～n |f i |

[0082]

[0083] Among them, f i w1 represents the i-th feature value in the set of all feature values ​​of the multi-scale semantic encoding vector of the first target task, and w1 represents the f i The corresponding first target task multi-scale semantic coding space structure value, w2 represents the f i The corresponding second objective task's multi-scale semantic coding space structure value;

[0084] Secondly, for each feature value of the first target task multi-scale semantic coding vector, the long-range dependency value of the first target task multi-scale semantic coding is obtained by subtracting the product of the feature value and the total number of feature values ​​of the first target task multi-scale semantic coding vector from the spatial structure value of the first target task multi-scale semantic coding, which is expressed by the formula:

[0085] x i =w1-f i ×n

[0086] Where, x i Indicates the f i The corresponding long-range dependency value of the multi-scale semantic encoding of the first objective task, where n represents the total number of feature values ​​of the multi-scale semantic encoding vector of the first objective task;

[0087] Then, the long-range dependency value of the second target task multi-scale semantic encoding is obtained by multiplying the square root of the total number of feature values ​​by the product of the feature values ​​and subtracting the spatial structure value of the second target task multi-scale semantic encoding. This value is expressed by the formula:

[0088]

[0089] Among them, y i Indicates the f i The corresponding second objective task is to encode long-range dependency values ​​using multi-scale semantic encoding.

[0090] Next, the exponent value calculated by using the exponent of the long-range dependency value of the multi-scale semantic encoding of the first target task as the exponent of the natural constant is weighted and summed with the reciprocal of the long-range dependency value of the multi-scale semantic encoding of the second target task to obtain the optimized feature value corresponding to each feature value, expressed by the formula:

[0091]

[0092] Among them, f' i Indicates the f i The corresponding optimized feature values, α and β, represent different weight parameters, (·) -1 e represents the reciprocal of the expression. (·) This indicates the calculation of the exponent value with the natural constant as the base;

[0093] Finally, the optimized first target task multi-scale semantic encoding vector composed of the optimized feature values ​​is passed through a classifier-based task recognizer to obtain the urgency label and importance label of the first target task.

[0094] Here, when the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector represent the inter-word semantic association encoding features and context semantic encoding features of the first target task text parsing result, respectively, when performing feature attention interaction based on the dynamic memory mechanism, the inconsistency in the granularity of the text semantic features between the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector will cause attention interaction mismatch, resulting in the first target task multi-scale semantic encoding vector having differences in the multi-scale fusion distribution space structure of text semantic features, affecting the convergence consistency of classification and regression, and thus affecting the accuracy of the urgency label and importance label of the first target task obtained by the task recognizer based on the classifier.

[0095] Based on this, this application addresses the potential spatial structure gaps in the feature set of the first target task multi-scale semantic encoding vector in high-dimensional space, which can lead to inconsistent convergence of spatial structure information inference based on implicit features during classification and regression. This is achieved by establishing long-distance feature dependencies relative to the spatial structure representation of the first target task multi-scale semantic encoding vector based on the overall feature scale, thereby establishing the feature local connectivity of the first target task multi-scale semantic encoding vector. Furthermore, it captures spatial ambiguity information of object feature values ​​through unstructured feature point prediction of the first target task multi-scale semantic encoding vector. This enhances the spatial inductive bias perception capability of the feature set of the first target task multi-scale semantic encoding vector, improves the convergence consistency of classification and regression, and increases the accuracy of the urgency and importance labels of the first target task obtained by the task recognizer based on the classifier.

[0096] In the aforementioned AI-based time-limited early warning monitoring and supervision method, step S5 sets the early warning level and frequency based on the urgency and importance tags of the first target task. For example, if the first target task is marked as high urgency and high importance, the highest early warning level is configured for it, and early warning supervision information is sent at a high frequency. In a specific example of this application, the highest early warning level is set to issue an early warning 7 days before the deadline, and the early warning frequency is once every 2 hours to ensure that critical tasks receive timely attention and processing. Conversely, if the task's urgency and importance are low, the early warning level can be set to a regular early warning, and the early warning frequency can be appropriately reduced.

[0097] In the aforementioned AI-based time-limit warning monitoring and supervision method, step S6 involves monitoring and supervising the first target task based on a comparison between the current date and the deadline, and based on the warning level and frequency. Specifically, based on the comparison between the current date and the deadline, when the time remaining until the task deadline reaches a set warning level, a monitoring and supervision mechanism is automatically triggered, sending warning information to relevant responsible persons at a set frequency to monitor and supervise the first target task. This approach effectively balances resource allocation, ensuring that critical tasks are prioritized while avoiding excessive attention to low-priority tasks.

[0098] In summary, the AI-based time-limit warning monitoring and supervision method based on the embodiments of this application is explained. It first extracts target tasks from a task database and uses natural language processing technology to parse the target tasks, extracting the task type, deadline, and responsible person information. It further determines the urgency and importance of the task. Then, based on the urgency and importance, it sets the warning level and frequency for the target task. When the time remaining until the task deadline reaches the set warning level, the monitoring and supervision mechanism is automatically triggered, sending warning information to the relevant responsible persons at the set frequency to check and remind them of the task progress. In this way, the efficiency and effectiveness of supply chain task management can be effectively improved, reducing losses caused by task delays.

[0099] Furthermore, an AI-based time-limited early warning monitoring and supervision system is also provided.

[0100] Figure 6 This is a block diagram of an AI-based time-limit early warning monitoring and supervision system according to an embodiment of this application. Figure 6 As shown, the AI-based time-limit warning monitoring and supervision system 100 according to an embodiment of this application includes: a first target task extraction module 110, used to extract a first target task from a task library; a text parsing module 120, used to parse the first target task using natural language processing technology to obtain a text parsing result of the first target task; a key information extraction module 130, used to extract key information of the first target task from the text parsing result of the first target task, the key information including task type, deadline, and person in charge; a priority evaluation module 140, used to determine the urgency label and importance label of the first target task based on the text parsing result of the first target task; a warning level setting module 150, used to set a warning level and frequency based on the urgency label and importance label of the first target task; and a task monitoring and supervision module 160, used to monitor and supervise the first target task based on a comparison between the current date and the deadline, and based on the warning level and frequency.

[0101] Those skilled in the art will understand that the specific operations of each module in the AI-based time-limited early warning monitoring and supervision system have been referenced above. Figures 1 to 5 The description of the AI-based time-limited early warning monitoring and supervision method is detailed here, and therefore, its repeated description will be omitted.

[0102] The basic principles of the present invention have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in the present invention are merely examples and not limitations, and should not be considered as essential features of each embodiment of the present invention. Furthermore, the specific details of the above embodiments are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the present invention to the necessity of employing the specific details described above.

[0103] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the unit division is only a logical functional division, and other division methods may exist in actual implementation. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0104] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the present invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0105] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in a system claim may also be implemented by a single unit through software or hardware.

[0106] Finally, it should be noted that the above description has been given for illustrative and descriptive purposes. Furthermore, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention.< / deadline>

Claims

1. A time-limited early warning monitoring and supervision method based on AI technology, characterized in that, include: Retrieve the first target task from the task library; Natural language processing techniques are used to parse the first target task to obtain the text parsing result of the first target task; Extract key information of the first target task from the text parsing result of the first target task, including task type, deadline, and person in charge; Based on the text parsing results of the first target task, determine the urgency and importance tags of the first target task; Based on the urgency and importance tags of the first target task, set the warning level and frequency; Based on the comparison between the current date and the deadline, and based on the warning level and frequency, the first target task is monitored and supervised.

2. The time-limited early warning monitoring and supervision method based on AI technology according to claim 1, characterized in that, Based on the text parsing results of the first target task, the urgency and importance tags of the first target task are determined, including: The text parsing results of the first target task are subjected to word-granular semantic embedding encoding to obtain a sequence of word-granular semantic embedding encoding vectors for the first target task; The sequence of the first target task word-granular semantic embedding encoding vectors is subjected to inter-word local semantic association encoding and global context semantic association encoding to obtain the first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector. Fine-grained interactive fusion is performed on the semantic association encoding vector between words of the first target task and the semantic encoding vector of the context of the first target task to obtain the multi-scale semantic encoding vector of the first target task; Based on the multi-scale semantic encoding vector of the first target task, the urgency label and importance label of the first target task are determined.

3. The time-limited early warning monitoring and supervision method based on AI technology according to claim 2, characterized in that, The first target task text parsing result is subjected to word-granular semantic embedding encoding to obtain a sequence of first target task word-granular semantic embedding encoding vectors, including: The text parsing result of the first target task is processed by word segmentation to obtain the text distribution sequence of the first target task word granularity description; Each first target task word granularity description in the text distribution sequence of the first target task word granularity description is input into a word embedding encoder based on the BERT model to obtain a sequence of semantic embedding encoding vectors for the first target task word granularity.

4. The time-limited early warning monitoring and supervision method based on AI technology according to claim 3, characterized in that, The sequence of the first target task word-granular semantic embedding encoding vectors is subjected to inter-word local semantic association encoding and global contextual semantic association encoding to obtain the first target task inter-word semantic association encoding vector and the first target task contextual semantic encoding vector, including: The sequence of semantic embedding encoding vectors of the first target task word granularity is input into the inter-word semantic information extraction network based on a one-dimensional convolutional layer to obtain the inter-word semantic association encoding vector of the first target task. The sequence of the first target task word-granular semantic embedding encoding vector is input into the global context semantic encoder based on the converter to obtain the first target task context semantic encoding vector.

5. The time-limited early warning monitoring and supervision method based on AI technology according to claim 4, characterized in that, Fine-grained interactive fusion is performed on the inter-word semantic association encoding vector of the first target task and the context semantic encoding vector of the first target task to obtain the multi-scale semantic encoding vector of the first target task, including: The first target task word semantic association encoding vector and the first target task context semantic encoding vector are respectively optimized by feature modulation based on external knowledge to obtain the optimized first target task word semantic association encoding vector and the optimized first target task context semantic encoding vector. The optimized first target task word semantic association encoding vector and the optimized first target task context semantic encoding vector are subjected to position-by-position semantic interactive encoding to obtain the first target task multi-scale semantic encoding vector.

6. The time-limited early warning monitoring and supervision method based on AI technology according to claim 5, characterized in that, The first target task inter-word semantic association encoding vector and the first target task context semantic encoding vector are respectively subjected to feature modulation optimization based on external knowledge to obtain the optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector, including: The semantic association encoding vector between words of the first target task and the semantic encoding vector of the context of the first target task are input into the fine-grained feature interaction network to obtain the fine-grained interaction matrix of multi-scale semantic features of the first target task; The fine-grained interaction matrix of the multi-scale semantic features of the first target task is input into the attention unit based on external knowledge to obtain the fine-grained interaction matrix of the multi-scale semantic features of the first target task optimized by external knowledge. Based on the external knowledge, optimize the fine-grained interaction matrix of multi-scale semantic features of the first target task, and perform feature modulation optimization on the word semantic association encoding vector and the context semantic encoding vector of the first target task to obtain the optimized word semantic association encoding vector and the optimized context semantic encoding vector of the first target task.

7. The time-limited early warning monitoring and supervision method based on AI technology according to claim 6, characterized in that, Based on the external knowledge, optimize the fine-grained interaction matrix of multi-scale semantic features for the first target task. Perform feature modulation optimization on the inter-word semantic association encoding vector and the context semantic encoding vector of the first target task respectively to obtain the optimized inter-word semantic association encoding vector and the optimized context semantic encoding vector of the first target task, including: A linear transformation is performed on the semantic association encoding vector between words of the first target task to obtain a first query feature vector and a first value feature vector. The external knowledge is used to optimize the fine-grained interaction matrix of the multi-scale semantic features of the first target task as a key matrix. The first query feature vector, the first value feature vector and the key matrix are input into a fine-grained modulation module based on the Transformer structure to obtain the optimized semantic association encoding vector between words of the first target task. A linear transformation is performed on the first target task context semantic encoding vector to obtain a second query feature vector and a second value feature vector. The external knowledge is used to optimize the fine-grained interaction matrix of the first target task multi-scale semantic features as the key matrix. The second query feature vector, the second value feature vector, and the key matrix are input into the fine-grained modulation module based on the Transformer structure to obtain the optimized first target task context semantic encoding vector.

8. The time-limited early warning monitoring and supervision method based on AI technology according to claim 7, characterized in that, The optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector are subjected to position-wise semantic interactive encoding to obtain the first target task multi-scale semantic encoding vector, including: The first target task multi-scale semantic encoding vector is obtained by dividing the optimized first target task inter-word semantic association encoding vector and the optimized first target task context semantic encoding vector by the position point.

9. The time-limited early warning monitoring and supervision method based on AI technology according to claim 8, characterized in that, Based on the multi-scale semantic encoding vector of the first target task, the urgency label and importance label of the first target task are determined, including: The first target task multi-scale semantic encoding vector is input into a classifier-based task recognizer to obtain the urgency label and importance label of the first target task.

10. A time-limited early warning monitoring and supervision system based on AI technology, characterized in that, include: The first target task extraction module is used to extract the first target task from the task library; The text parsing module is used to parse the first target task using natural language processing technology to obtain the text parsing result of the first target task; The key information extraction module is used to extract key information of the first target task from the text parsing result of the first target task. The key information includes task type, deadline, and person in charge. The priority assessment module is used to determine the urgency and importance tags of the first target task based on the text parsing results of the first target task. The warning level setting module is used to set the warning level and frequency based on the urgency and importance tags of the first target task; The task monitoring and supervision module is used to monitor and supervise the first target task based on the comparison between the current date and the deadline, and based on the warning level and frequency.

Images