Method and device for intelligently allocating agent to event

Abstract

The invention relates to the technical field of natural language processing, and particularly provides a method and a device for intelligently allocating agents for events, based on a large government affair model, the method comprises the following steps: S1, constructing the large government affair model; s2, intelligent identification of events; s3, event classification; s4, making a rule; s5, intelligent Agent allocation is realized; s6, testing and optimizing; and S7, setting a feedback mechanism. Compared with the prior art, the problems of information processing lag, unreasonable resource allocation and the like in urban grassroots governance can be solved, modernization of urban governance is promoted, and a more efficient and more intelligent management and service mode is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and specifically provides a method and device for intelligent event dispatching agent. Background Art

[0002] The government affairs large model based on the gating network MOE can process complex government affairs data and extract valuable information from it. The government affairs large model usually requires a large amount of historical data and professional knowledge for training to improve the accuracy of its prediction and decision-making.

[0003] Events in urban grass-roots governance usually have multiple attributes, such as event type, occurrence location, involved departments, urgency level, etc. Analyzing these attributes can help determine the nature of the event and the priority of handling. Preliminary dispatching judgments can be made according to the attributes and criteria of the event. The rule engine can quickly process common events and ensure the consistency of processing.

[0004] The government affairs large model usually adopts various machine learning algorithms, such as random forest, support vector machine, neural network, etc., to train the model and predict the dispatching path of the event. Combining the prediction results of the rule engine and the government affairs large model for comprehensive decision-making. This method can combine the empirical knowledge of the rules and the prediction ability of the model to obtain more accurate dispatching decisions. Events in urban grass-roots governance often occur in real time, so technologies that can quickly process and analyze real-time data are needed. To facilitate user use, an intuitive and easy-to-use user interface needs to be designed, and effective user interaction methods, such as speech recognition, natural language processing, etc., need to be provided.

[0005] How to solve the problems of lagging information processing and unreasonable resource allocation in urban grass-roots governance in the prior art, promote the modernization of urban governance, and realize a more efficient and intelligent management and service mode are technical problems that need to be solved urgently by those skilled in the art. Summary of the Invention

[0006] The present invention aims at the deficiencies of the above-mentioned prior art and provides a method for intelligent event dispatching agent with strong practicability.

[0007] A further technical task of the present invention is to provide a device for intelligent event dispatching agent with reasonable design, safety and applicability.

[0008] The technical solution adopted by the present invention to solve its technical problems is:

[0009] A method for intelligent event dispatching agent, based on a government affairs large model, has the following steps:

[0010] S1. Construct a government affairs large model;

[0011] S2. Intelligent recognition of events;

[0012] S3. Event classification;

[0013] S4. Rule formulation;

[0014] S5. Implement intelligent dispatching Agent;

[0015] S6. Testing and optimization;

[0016] S7. Set up a feedback mechanism.

[0017] Furthermore, in step S1, four knowledge bases are constructed, namely the event aggregation library, the event transfer library, the business knowledge base, and the event storage library. Then, machine learning methods are used to train a model that can understand knowledge in the government affairs field. Finally, fine-tuning is performed based on the pre-trained language models BERT and XLNet;

[0018] The event aggregation library collects and stores event types from different channels and types;

[0019] The event transfer library records the status and transfer history of events throughout the processing flow;

[0020] The business knowledge base stores business rules, standards, and knowledge related to event processing, guiding event classification, dispatching, and handling;

[0021] The event storage library is used to store event data in the long term;

[0022] Furthermore, in step S2, event intelligent recognition includes text parsing, construction of a knowledge graph, event standardization, event distribution, and entity recognition;

[0023] The text parsing is to use NLP technology to parse the event description and extract key information;

[0024] The construction of the knowledge graph is to build a dynamic knowledge graph by sorting out the knowledge system in the government affairs field;

[0025] The event standardization is to define a set of standard event classification and coding schemes;

[0026] The event distribution is to distribute the event to the corresponding processing department according to the standard category of the event;

[0027] The entity recognition is to construct an event dispatching knowledge base, which includes data integration, interactive query, and knowledge extraction, to accurately identify key entities in the event.

[0028] Further, in step S3, according to the event description, classify the event type, input the collected event description or report into the government affairs large model, preprocess the text using NLP technology, extract the feature vectors of the text according to the pre-trained word vector model, and input the feature vectors into the trained government affairs large model. The government affairs large model will output the category prediction of the event.

[0029] Further, in step S4, the rule formulation includes event type rules, urgency rules, location-based distribution rules, and composite rule formulation. Then, weights are assigned to different event features to determine the distribution priority.

[0030] The event type rules are to list the event types and formulate distribution criteria for each type.

[0031] The urgency rules are to define the urgency levels, determine the processing priorities according to the urgency of the events, and formulate corresponding distribution logics:

[0032] The location-based distribution rules are designed by combining multiple factors and comprehensive rules.

[0033] Further, when assigning weights to different event features and determining the distribution priority, the following steps are involved:

[0034] (1) Determine the feature importance, identify the key features, and analyze the influence degree of each feature on the event processing result.

[0035] (2) Set the weight range and determine the weight interval. Usually, the weights will be assigned between 0 and 1, where 0 means the feature has no influence on the priority, and 1 means the feature has a decisive influence on the priority. Assign a preliminary weight to each feature according to its importance.

[0036] (3) Solicit opinions from domain experts to adjust the preliminary weights; adjust the weights again through historical data analysis to better reflect the actual situation.

[0037] (4) After comprehensively considering the expert opinions and data analysis results, determine the final weight of each feature.

[0038] (5) Formulate weight rules to convert the weight assignment into operable rules. The features of the event are event type, urgency, and influence scope.

[0039] The comprehensive weight calculation formula is: Comprehensive weight = (Event type weight + Urgency weight + Influence scope weight) / 3;

[0040] According to the comprehensive weight, determine the distribution priority of the event. The higher the weight, the higher the priority.

[0041] Further, in step S5, the trained government affairs large model is integrated into the distribution system. Based on the distribution rules of the decision tree, distribution is performed, and the events are distributed to the corresponding processing paths. According to the preset rules and standards, preliminary distribution judgment is made on the events. The basic structure and parameters of the rules are defined as follows:

[0042] If the event type T is "consultation" and the department involved D is "customer service", then the distribution path P is "Path A";

[0043] If the event type T is "complaint" and the urgency U is "high", then the distribution path P is "Path B";

[0044] The rule is expressed as a mathematical formula:

[0045]

[0046] Use the government affairs large model to conduct in-depth analysis of the events and predict the most suitable distribution path.

[0047] Further, when using the government affairs large model to conduct in-depth analysis of the events and predict the most suitable distribution path, assume there is an event E that has multiple attributes. The government affairs large model takes these attributes as inputs and outputs a distribution path P;

[0048] The inference function of the government affairs large model is:

[0049] P(E) = f(T, U, D);

[0050] Among them, f is a function that predicts the distribution path P based on the event type T, urgency U, and department involved D;

[0051] Combine the rules and the results of the government affairs large model for comprehensive decision-making, and dock with the information system API of relevant departments to achieve real-time data interaction.

[0052] Further, in step S6, prepare a test data set to test the intelligent distribution Agent, evaluate the accuracy, speed, and system stability of the distribution, and optimize the model and distribution logic according to the test results;

[0053] In step 7, collect the processing results and feedback of the responsible department, and use the feedback data to further optimize the model and distribution rules.

[0054] A device for an event intelligent distribution agent includes: at least one memory and at least one processor;

[0055] The at least one memory is used to store machine-readable programs;

[0056] The at least one processor is configured to call the machine-readable program to execute a method for an event intelligent dispatching agent.

[0057] Compared with the prior art, a method and device for an event intelligent dispatching agent of the present invention have the following prominent beneficial effects:

[0058] The present invention can not only improve the efficiency and quality of urban grass-roots governance, but also provide important support for building a smart city and enhancing the service capabilities of relevant departments.

[0059] Automated event classification and dispatching processes can significantly reduce manual processing time and improve event response speed. By leveraging the deep learning capabilities of the government affairs large model, event types and key information can be identified more accurately, reducing misclassification cases.

[0060] The intelligent dispatching system can reasonably allocate resources of relevant departments and staff according to the urgency and type of events, improving resource utilization efficiency. By analyzing event data, decision-making support can be provided for relevant departments to help formulate more effective strategies and plans.

[0061] Quickly responding to and handling citizens' demands can enhance citizens' satisfaction with department services and strengthen the interaction between departments and citizens. Building a government affairs large model requires cross-departmental data integration, which helps promote data sharing and collaboration between departments. Automated event processing processes can reduce reliance on human resources, thereby reducing long-term operating costs. The intelligent dispatching system can record every step of event processing, increasing work transparency and facilitating supervision and accountability.

[0062] The dispatching Agent can continuously learn from new events, optimize model performance, and adapt to the changing urban governance needs. In the event of an emergency, the intelligent dispatching system can quickly identify and dispatch, accelerating the emergency response speed and reducing possible losses. Applying advanced machine learning and artificial intelligence technologies can promote innovation and development in the application of technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0064] Attached Figure 1 is a flowchart showing a method for an event intelligent dispatching agent. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] To enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below in conjunction with specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0066] The following gives an optimal embodiment:

[0067] As Figure 1 shown, a method for an event intelligent dispatching agent in this embodiment, based on a government affairs large model, has the following steps:

[0068] S1. Construct a government affairs large model;

[0069] Data collection: Collect a large amount of government affairs data, including historical event processing records, policy documents, department responsibility descriptions, etc.

[0070] Construct four knowledge bases:

[0071] Event aggregation library: This library is the starting point of the process, used to collect and store event information from different channels and types. It is a centralized event information storage center. It may include basic information of events, such as event ID, reporting time, reporter information, event description, etc. Ensure that all events can be recorded and tracked, providing raw data for subsequent processing.

[0072] Event transfer library: This library records the status and transfer history of events in the entire processing process. It tracks every step of the event from reporting to dispatching, processing, and feedback. It may include event status, processing department, processing personnel, processing time, processing result, transfer record, etc. Provide transparency of event processing, facilitating supervision and tracing of the event processing process.

[0073] Business knowledge base: This library stores business rules, standards, and knowledge related to event processing. It is used to guide the classification, dispatching, and disposal of events. It may include event type definitions, emergency level standards, responsibilities of responsible departments, processing process guidelines, relevant policies and regulations, etc. Provide decision-making support for event intelligent dispatching to ensure that events can be correctly processed according to established rules.

[0074] Event storage library: This library is used to store event data in the long term, including information on historical events and current events, facilitating data analysis and reporting. It may include detailed information on all events, including the original data of the event, processing process, processing result, relevant attachments, etc. Support data analysis and learning, help improve the event processing process, and improve the efficiency and accuracy of future event processing.

[0075] Model training: Train a model using machine learning methods that can understand knowledge in the government affairs domain. This can be fine-tuned based on pre-trained language models such as BERT, XLNet, etc.

[0076] S2. Intelligent identification of events;

[0077] Text parsing: Use NLP techniques to parse event descriptions and extract key information.

[0078] Knowledge graph construction: Build a dynamic knowledge graph by sorting out the knowledge system in the government affairs domain. This includes entity recognition, relation extraction, and knowledge fusion.

[0079] Event standardization: Define a set of standard event classification and coding schemes for the unified management and query of events.

[0080] Event distribution: Distribute events to the corresponding processing departments or systems according to the standard categories of events.

[0081] Entity recognition: When dealing with historical grass-roots governance events, building an event dispatching knowledge base is a crucial step. This knowledge base includes functions such as data integration, interactive query, and knowledge extraction, which can help the system more accurately identify key entities in events. Through data integration, the system can integrate data from different sources to ensure the integrity and accuracy of information. Interactive query allows users to flexibly search and filter information according to their needs, improving efficiency. Knowledge extraction can automatically extract key entities from text, such as locations, people, and organization names, laying a foundation for subsequent analysis and processing.

[0082] S3. Event classification;

[0083] Event classification: Classify event types according to event descriptions. Input the collected event descriptions or reports into the model. Use natural language processing (NLP) techniques to preprocess the text, including removing stop words, stemming, and part-of-speech tagging. Extract the feature vectors of the text according to the pre-trained word vector model, and these vectors can represent the semantic information of the event. Input the feature vectors into the trained model, and the model will output the category prediction of the event. The model may use multiple classification algorithms, such as gradient boosting decision tree (GBDT) or deep learning network. An event may belong to multiple categories at the same time, so the model needs to be able to handle multi-label classification problems. For the classification results, it may be necessary to set a threshold to determine whether to accept the prediction results of the model.

[0084] S4. Rule formulation;

[0085] Rule formulation: Formulate dispatching rules according to information such as event type, urgency, and geographical location.

[0086] Event type rules: List event types: environmental hygiene, public safety, traffic management, etc. Develop allocation criteria for each type: For example, all environmental hygiene events are allocated to the environmental hygiene department.

[0087] Urgency rules: Define urgency levels, and determine the processing priorities based on the urgency of the events. Develop corresponding allocation logics: For example, very urgent events need to be immediately allocated to the highest-level processing unit.

[0088] Geographical location rules: Divide geographical areas, based on administrative regions or geographical features.

[0089] Develop location-based allocation rules: For example, events occurring in a certain street are allocated to the management department of that street. Combine multiple factors in the comprehensive rule design, considering factors such as event type, urgency, and geographical location.

[0090] Develop composite rules: For example, if the event type is "traffic accident", the urgency is "urgent", and it occurs in the "city center", then it is allocated to the traffic police brigade and the ambulance.

[0091] Weight allocation: Allocate weights to different event characteristics to determine the allocation priorities;

[0092] Step 1: Determine feature importance, identify key features: Determine which event characteristics are most important for allocation priorities, such as event type, urgency, impact scope, geographical location, etc. Evaluate the feature impacts: Analyze the impact degree of each feature on the event processing results.

[0093] Step 2: Set weight ranges, determine the weight intervals: Usually, weights are allocated between 0 and 1, where 0 means the feature has no impact on the priority, and 1 means the feature has a decisive impact on the priority. Allocate preliminary weights, and assign a preliminary weight to each feature according to its importance.

[0094] Step 3: Weight adjustment, expert opinions: Solicit opinions from domain experts to adjust the preliminary weights. Data analysis: Adjust the weights through historical data analysis to better reflect the actual situation.

[0095] Step 4: Determine the weight allocation, final determination: After comprehensively considering expert opinions and data analysis results, determine the final weights of each feature.

[0096] Step 5: Write weight rules, develop weight rules: Convert the weight allocation into actionable rules

[0097] Suppose the characteristics of an event are as follows:

[0098] Event type: Public safety event (weight 0.5)

[0099] Urgency: Extremely urgent (weight 0.6)

[0100] Scope of impact: Wide impact (weight 0.4)

[0101] The comprehensive weight calculation formula can be: Comprehensive weight = (Event type weight + Urgency weight + Scope of impact weight) / 3;

[0102] Substitute the values: Comprehensive weight = (0.5 + 0.6 + 0.4) / 3 = 0.5. Based on the comprehensive weight, the system can determine the dispatching priority of the event. The higher the weight, the higher the priority. In this way, it can be ensured that the most urgent and important events are processed first.

[0103] S5. Implement an intelligent dispatching Agent;

[0104] Integration model: Integrate the trained government affairs large model into the dispatching system. Package the trained model into a service, which can be a RESTful API or a message queue service for easy invocation by the dispatching system. Integrate the model service in the dispatching system to ensure that the system can send event data to the model and receive the prediction results of the model.

[0105] Dispatching algorithm: Implement a dispatching algorithm based on rules and model prediction, such as using decision trees, support vector machines, or deep learning models. Dispatch according to the rules, implement the dispatching rules based on decision trees, and dispatch events to the corresponding processing paths through a series of judgment conditions. Set predefined rules according to attributes such as event type and urgency.

[0106] Intelligent decision-making: Rule matching: Make a preliminary dispatching judgment on the event according to the preset rules and standards. First, we need to define a set of preset rules and standards, which should be based on historical data and expert knowledge. These rules can be simple logical judgments or complex algorithms. In government affairs services, preliminary dispatching can be carried out according to factors such as event type, urgency, and involved departments. To design a mathematical formula to represent the rule matching process in intelligent decision-making, we first need to define the basic structure and parameters of the rules. For example, a simple rule can be:

[0107] If the event type T is "Consultation" and the involved department D is "Customer Service", then the dispatching path P is "Path A".

[0108] If the event type T is "Complaint" and the urgency U is "High", then the dispatching path P is "Path B".

[0109] We can represent these rules as mathematical formulas. For example:

[0110]

[0111] Use the government affairs large model to conduct in-depth analysis of events and predict the most appropriate distribution path. In the process of using the government affairs large model for in-depth analysis and predicting the most appropriate distribution path, we first need to define the input and output of the model. Suppose we have an event E, which has multiple attributes, such as event type T, urgency U, departments involved D, etc. These attributes are used as input, and a distribution path P is output.

[0112] The model inference function can be:

[0113] P(E) = f(T, U, D);

[0114] where f is a function that predicts the distribution path P based on the event type T, urgency U, and departments involved D;

[0115] f is a reinforcement learning model trained based on historical data, such as a neural network. The training process of this model involves learning from historical data how to predict the distribution path according to the attributes of the event. Combining the rules and the results of the model, comprehensive decisions are made. Connect to the information system API of relevant departments to achieve real-time data interaction.

[0116] S6. Testing and optimization;

[0117] Test dataset: Prepare a test dataset to test the intelligent distribution Agent.

[0118] Performance evaluation: Evaluate the accuracy, speed, and stability of the distribution.

[0119] Iterative optimization: Optimize the model and distribution logic according to the test results.

[0120] S7. Set up a feedback mechanism;

[0121] Collect the processing results and feedback from the responsible departments.

[0122] Use the feedback data to further optimize the model and distribution rules.

[0123] Based on the above method, a device for an event intelligent distribution agent in this embodiment includes: at least one memory and at least one processor;

[0124] The at least one memory is used to store machine-readable programs;

[0125] The at least one processor is used to call the machine-readable program and execute a method for an event intelligent distribution agent.

[0126] The above specific embodiments are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above specific embodiments. Any technical solution that conforms to the above specific embodiments of the present invention and any appropriate changes or substitutions made by any person of ordinary skill in the relevant technical field shall fall within the patent protection scope of the present invention.

[0127] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligently distributing an event to an agent, characterized in that: Based on the government affairs big model, there are the following steps: S1. Build a big government model; S2, intelligent identification of events; S3, event classification; S4. Rule making; S5, realize intelligent distribution Agent; S6. Testing and optimization; S7. Set up a feedback mechanism.

2. The method of intelligent event distribution agent according to claim 1, characterized in that: In step S1, four knowledge bases are constructed, namely, the event aggregation base, the event flow base, the business knowledge base, and the event storage base. Then, a machine learning method is used to train a model that can understand government domain knowledge. Finally, fine-tuning is performed based on the BERT and XLNet pre-trained language models. The event aggregation library collects and stores event types from different channels and types; The event flow library records the status and flow history of events in the entire processing flow; The business knowledge base stores business rules, standards and knowledge related to event processing, and guides the classification, allocation and handling of events; The event storage repository is used to store event data for a long period of time.

3. The method of intelligent event distribution agent according to claim 2, characterized in that: In step S2, event intelligent identification includes text parsing, knowledge graph construction, event standardization, event distribution, and entity recognition; The text parsing is to parse the event description using NLP technology to extract key information; The knowledge graph is constructed by combing the knowledge system in the government affairs field to construct a dynamic knowledge graph; The event standardization is to define a set of standard event classification and coding schemes; The event distribution is to distribute the event to corresponding processing departments according to the standard category of the event; The entity recognition is to build an event distribution knowledge base, which includes data integration, interactive query and knowledge extraction to accurately identify key entities in the event.

4. The method of intelligent event distribution agent according to claim 3, characterized in that: In step S3, the event type is classified according to the event description, and the collected event description or report is input into the government affairs big model. The text is preprocessed using NLP technology, and the feature vector of the text is extracted according to the pre-trained word vector model. The feature vector is input into the trained government affairs big model, and the government affairs big model will output the category prediction of the event.

5. The method of intelligent event distribution agent according to claim 4, characterized in that: In step S4, the rule formulation includes event type rules, urgency rules, geographic location-based allocation rules and compound rules, and then weights are assigned to different event features to determine allocation priorities; The event type rule is to list the event types and formulate allocation criteria for each type; The urgency rule is to define the urgency level, determine the processing priority according to the urgency of the event, and formulate the corresponding allocation logic: The allocation rule based on geographical location is designed by combining multiple factors.

6. The method of intelligent event distribution agent according to claim 5, characterized in that: The following steps are involved in assigning weights to different event features and determining allocation priorities: (1) Determine the importance of features, identify key features, and analyze the impact of each feature on event processing results; (2) Set the weight range and determine the weight interval. Usually, the weight is allocated between 0 and 1, where 0 means that the feature has no effect on the priority and 1 means that the feature has a decisive influence on the priority; Assign a preliminary weight to each feature based on its importance; (3) Seek opinions from experts in the field and adjust the initial weights; through historical data analysis, adjust the weights again to better reflect the actual situation; (4) After comprehensively considering expert opinions and data analysis results, determine the final weight of each feature; (5) Formulate weight rules to convert weight allocation into operational rules. The characteristics of events are event type, urgency, and scope of impact. The formula for calculating the comprehensive weight is: comprehensive weight = (event type weight + urgency weight + impact range weight) / 3; Determine the event allocation priority based on the comprehensive weight. The higher the weight, the higher the priority.

7. The method of intelligent event distribution agent according to claim 6, characterized in that: In step S5, the trained government affairs big model is integrated into the distribution system, and the distribution is performed based on the distribution rules of the decision tree, and the events are distributed to the corresponding processing paths. According to the preset rules and standards, the events are preliminarily distributed and judged. The basic structure and parameters of the rules are defined as follows: If the event type T is "consultation" and the department involved D is "customer service", the distribution path P is "path A"; If the event type T is "complaint" and the urgency U is "high", the distribution path P is "path B"; The rule is expressed as a mathematical formula: Use the government affairs big data model to conduct in-depth analysis of events and predict the most appropriate distribution path.

8. The method of intelligent event distribution agent according to claim 7, characterized in that: When using the government affairs big model to conduct in-depth analysis of events and predict the most appropriate distribution path, suppose there is an event E, which has multiple attributes. The government affairs big model takes these attributes as input and outputs a distribution path P; The inference function of the government affairs model is: P(E)=f(T,U,D); Where f is a function that predicts the distribution path P based on the event type T, urgency U, and involved departments D; Combine the results of rules and government affairs big models to make comprehensive decisions, connect with the information system API of relevant departments, and realize real-time data interaction.

9. The method of intelligent event distribution agent according to claim 8, characterized in that: In step S6, a test data set is prepared to test the intelligent distribution agent, evaluate the distribution accuracy, speed and system stability, and optimize the model and distribution logic according to the test results; In step 7, the processing results and feedback from the responsible departments are collected, and the feedback data is used to further optimize the model and allocation rules.

10. A device for intelligent event distribution agent, characterized in that: include: at least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to call the machine-readable program to execute the method according to any one of claims 1 to 9.

Images