Government affair hotline work order distribution method and system based on large language model

Abstract

The invention discloses a government affair hotline work order distribution method and system based on a large language model, relates to the technical field of intelligent government affairs, and solves the problems that the prior art often depends on manual judgment and experience, and meanwhile, whether a to-be-distributed department can process a work order within effective time is lacked of consideration. And the technical problem of low accuracy and efficiency of the work order distribution system is solved. Work order data is generated according to hotline data; constructing a distribution objective function according to the work order data, and then solving to obtain a distribution suggestion; distributing the work order data according to a distribution suggestion; the processing duration of work order data is acquired in real time, an alarm signal is generated according to the processing duration, dialect characteristics of various regions are considered, so that the work order distribution system can accurately identify appeals of different crowds, meanwhile, a distribution objective function is constructed, and work orders can be timely and effectively distributed by comprehensively considering multiple factors; and the accuracy and efficiency of the work order distribution system are improved.

Description

Technical Field

[0001] This application belongs to the field of intelligent government technology, specifically a method and system for dispatching government service hotline work orders based on large language models. Background Art

[0002] Large language models are advanced natural language processing tools based on deep learning technology and trained using massive text data. They can capture complex patterns of language, understand context, and generate coherent and meaningful text. These models have demonstrated powerful capabilities in multiple fields such as text generation, translation, and question answering, greatly promoting the development of natural language processing technology. Despite challenges such as data privacy and interpretability, large language models still have broad development prospects.

[0003] The government service hotline system is a comprehensive work platform that integrates various system functions and implements "one-stop" centralized handling and services. With the continuous deepening of the service-oriented government, the volume of government service hotline work orders has increased sharply, putting great pressure on manual customer service. Moreover, the work order matters are complex, involving many departments, making it difficult to identify difficult work orders and the accuracy of order dispatching is low, thus affecting the work order handling time and public satisfaction. Existing work order dispatching systems often rely on manual judgment and experience, and at the same time lack consideration of whether the department to which the work order is to be dispatched can process the work order within an effective time, resulting in low accuracy and efficiency of the work order dispatching system; therefore, the government service hotline work order dispatching system still needs further improvement. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art; for this purpose, this application proposes a method and system for dispatching government service hotline work orders based on large language models, which are used to solve the technical problems that existing technologies often rely on manual judgment and experience, and at the same time lack consideration of whether the department to which the work order is to be dispatched can process the work order within an effective time, resulting in low accuracy and efficiency of the work order dispatching system.

[0005] To achieve the above object, the first aspect of this application provides a government service hotline work order dispatching system based on large language models, including: a data collection module, a data analysis module, an early warning module, and a database;

[0006] The data collection module: obtains hotline data through data collection devices;

[0007] The data analysis module: generates work order data based on the hotline data; constructs a dispatching objective function according to the work order data and then solves to obtain a dispatching suggestion; dispatches the work order data according to the dispatching suggestion; obtains the processing duration of the work order data in real time, and generates an alarm signal according to the processing duration;

[0008] The early warning module: makes a prompt according to the alarm signal and contacts the management personnel;

[0009] The database is used to store historical data required for training the model.

[0010] Through the above steps, this application incorporates the dialect characteristics of different regions, ensures that the work order dispatching system can accurately capture and understand the needs of different groups, and constructs a dispatching objective function that comprehensively integrates multiple consideration factors, aiming to achieve efficient and timely allocation of work orders, and improves the accuracy and operation efficiency of the work order dispatching system.

[0011] Further, generating work order data based on the hotline data includes:

[0012] Obtaining hotline data; the hotline data includes text data and voice data;

[0013] Judging whether the hotline data is voice data;

[0014] If yes, inputting the voice data into a speech recognition model to obtain speech-text data and speech-text labels; the speech recognition model is constructed through an artificial intelligence model;

[0015] If no, inputting the text data or speech-text data and their corresponding speech-text labels into a large language model to obtain work order data; the large language model is obtained by fine-tuning on government affairs domain data.

[0016] Further, the speech recognition model is constructed through an artificial intelligence model, including:

[0017] Obtaining a number of historical voice data and their corresponding historical speech-text data and speech-text labels;

[0018] Dividing a number of historical voice data and their corresponding historical speech-text data and speech-text labels into training data, validation data, and test data; performing data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set;

[0019] Selecting an artificial intelligence model as the basic model;

[0020] Training the basic model with the training set, and adjusting the learning rate and hyperparameters on the validation set to obtain a pre-trained model;

[0021] Verifying the pre-trained model on the test set, and finally obtaining a speech recognition model that inputs voice data and outputs speech-text data and speech-text labels.

[0022] Further, the large language model is obtained by fine-tuning on government affairs domain data, including:

[0023] Obtain a pre-trained large language model and government domain data; the government domain data includes a number of historical text data and their corresponding work order data, audio-text data, audio-text labels and their corresponding work order data;

[0024] Divide a number of historical text data and their corresponding work order data, audio-text data, audio-text labels and their corresponding work order data into training data, validation data and test data; perform data preprocessing on the training data, validation data and test data to obtain a training set, a validation set and a test set;

[0025] Train the pre-trained large language model with the training set, and adjust the learning rate and hyperparameters on the validation set;

[0026] Verify the pre-trained large language model on the test set, and finally obtain a large language model that takes input text data or audio-text data and their corresponding audio-text labels and outputs work order data.

[0027] Further, after constructing a dispatch objective function based on the work order data and solving to obtain a dispatch suggestion, it includes:

[0028] Obtain work order data; the work order data includes consultation time, work order location, deadline duration, work order content, content level, and the corresponding confidence level ZD of the department to which it belongs;

[0029] Generate a number of target quantization functions based on the deadline duration, content level, and the corresponding confidence level of the department to which it belongs;

[0030] Through the formula Construct a dispatch objective function FC; where, ω1, ω2 and ω3 are dynamic weight coefficients, ω1∈(α1, α2), ω2∈(α3, α4), ω3∈(α5, α6); the dynamic weight coefficients are generated according to the work order data; α1, α2, α3, α4, α5 and α6 are constants, and α1, α2, α3, α4, α5 and α6∈(0, 1); DJ is the maximized level quantization function, DD is the unit level; QT is the maximized deadline occupancy ratio quantization function, DT is the unit time ratio; Z is the maximized confidence level quantization function, DZ is the unit confidence level; BRL is the maximized department task completion rate, DL is the unit completion rate;

[0031] Generate a dispatch suggestion according to the dispatch objective function.

[0032] Further, generating a number of target quantization functions based on the deadline duration, content level, and the corresponding confidence level of the department to which it belongs includes:

[0033] Obtain the work order pending dispatch time in real time, and obtain the consultation time, deadline duration QS, content level, the corresponding confidence level of the department to which it belongs, and the department completion rate BWL;

[0034] Construct the maximized level quantization function ND through the formula ND = max{DJ i}; where DJ i represents the content level of the i-th work order data;

[0035] Obtain the waiting duration DS by taking the difference between the work order pending dispatch time and the consultation time;

[0036] Construct the minimized deadline occupancy ratio quantization function QT through the formula ; where DS i represents the waiting duration of the i-th work order data; QS i represents the deadline duration of the i-th work order data;

[0037] Construct the maximized confidence quantization function Z through the formula Z = max{ZD ij}; where ZD ij represents the j-th confidence in the i-th work order data;

[0038] Construct the maximized department task completion rate BRL through the formula BRL = max{BWL j}; where BWL j represents the department completion rate of the department corresponding to the j-th confidence.

[0039] Furthermore, the dynamic weight coefficient is generated according to the work order data, including:

[0040] Obtain the work order data; the work order data includes the deadline duration, content level, and the department and its corresponding confidence it belongs to;

[0041] Input the work order data into the weight prediction model to obtain the corresponding dynamic weight coefficient;

[0042] Among them, the weight prediction model is constructed through a machine learning model, including:

[0043] Obtain a number of historical work order data and their corresponding historical weight coefficients;

[0044] Divide a number of historical work order data and their corresponding historical weight coefficients into training data, validation data, and test data; perform data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set;

[0045] Select a machine learning model as the basic model;

[0046] Train the basic model through the training set, and adjust the learning rate and hyperparameters on the validation set to obtain a pre-trained model;

[0047] By validating the pre-trained model on the test set, the weight prediction model that takes the input work order data and outputs several dynamic weight coefficients is finally obtained.

[0048] Through the above steps, the present application updates the dynamic weight coefficients in real time, enabling the system to adaptively adjust the appropriate weight coefficients under different work order data, thereby improving the accuracy and efficiency of the work order dispatching system.

[0049] Further, generating the dispatching suggestion according to the dispatching objective function includes:

[0050] Obtain the dispatching objective function;

[0051] Obtain the optimal solution of the dispatching objective function by passing the production data and the product comprehensive function through the deep reinforcement learning model, and generate the dispatching suggestion based on this.

[0052] Further, generating the alarm signal according to the processing duration includes:

[0053] Obtain the processing duration and the deadline duration of the work order data in real time;

[0054] Judge whether the processing duration is less than the deadline duration;

[0055] If yes, judge whether the processing duration is greater than D times the deadline duration; if yes, generate a warning signal that the processing is about to exceed the deadline; if not, do nothing;

[0056] If not, generate an alarm signal that the processing has exceeded the deadline; where D is a proportionality coefficient, D ∈ (0, 1).

[0057] Another aspect of the present invention provides a government hotline work order dispatching method based on a large language model, including:

[0058] S0: Obtain the hotline data;

[0059] S1: Generate work order data according to the hotline data;

[0060] S2: After constructing the dispatching objective function according to the work order data and solving it to obtain the dispatching suggestion; dispatch the work order data according to the dispatching suggestion;

[0061] S3: Obtain the processing duration of the work order data in real time, and generate an alarm signal according to the processing duration;

[0062] S4: Make a prompt according to the alarm signal and contact the management personnel.

[0063] Compared with the prior art, the beneficial effects of the present application are:

[0064] 1. This application generates work order data based on hotline data, constructs a dispatch objective function based on the work order data and then solves it to obtain dispatch suggestions, dispatches the work order data according to the dispatch suggestions, and obtains the processing duration of the work order data in real time, generates an alarm signal based on the processing duration, taking into account the dialect characteristics of each place, enabling the work order dispatch system to accurately identify the demands of different groups of people. At the same time, by constructing a dispatch objective function and comprehensively considering multiple factors, the work order can be dispatched in a timely and effective manner, improving the accuracy and efficiency of the work order dispatch system.

[0065] 2. This application constructs a number of objective quantification functions and constructs a dispatch objective function based on them, enabling the system to consider various aspects such as the time, level of the work order, and confidence of the department to which it belongs, and considering the task completion rate of the department on the basis of the department to which it belongs, so as to comprehensively consider and select the dispatched work order data and the dispatch department, improving the intelligence level and dispatch efficiency of the work order dispatch system. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application 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 only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0067] Figure 1 Schematic diagram of the principle of the government affairs hotline work order dispatch system based on the large language model of the present application;

[0068] Figure 2 Flowchart of the government affairs hotline work order dispatch method based on the large language model of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] The following will clearly and completely describe the technical solutions of the present application in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0070] Please refer to Figure 1 , the first aspect embodiment of the present application provides a government affairs hotline work order dispatch system based on the large language model, including: a data collection module, a data analysis module, an early warning module, and a database;

[0071] Data collection module: Obtain hotline data through data collection devices; the data collection devices include various sensors, etc.;

[0072] Data analysis module: Generate work order data based on hotline data; Construct a dispatch objective function based on the work order data and solve it to obtain dispatch suggestions; Dispatch the work order data according to the dispatch suggestions; Real-time obtain the processing duration of the work order data and generate an alarm signal based on the processing duration;

[0073] Early warning module: Make a prompt according to the alarm signal and contact the management personnel; The alarm signal includes an early warning signal for approaching expiration of processing and an alarm signal for expired processing, etc.;

[0074] The database is used to store historical data required for training the model.

[0075] Generating work order data according to the hotline data in this embodiment includes:

[0076] Obtain hotline data; The hotline data includes text data and voice data; The hotline data refers to the demands of enthusiastic people obtained through data collection devices; The voice data refers to the demands of enthusiastic people expressed in voice, and the text data refers to the demands of enthusiastic people expressed in text;

[0077] Judge whether the hotline data is voice data;

[0078] Yes, input the voice data into the speech recognition model to obtain speech-to-text data and speech-to-text labels; The speech recognition model is constructed through an artificial intelligence model; The speech-to-text data refers to the data obtained by converting speech into text, and the speech-to-text label refers to the dialect label corresponding to the voice data;

[0079] No, input the text data or speech-to-text data and their corresponding speech-to-text labels into the large language model to obtain work order data; The large language model is obtained by fine-tuning on government affairs domain data.

[0080] The speech recognition model in this embodiment is constructed through an artificial intelligence model, including:

[0081] Obtain a number of historical voice data and their corresponding historical speech-to-text data and speech-to-text labels;

[0082] Divide a number of historical voice data and their corresponding historical speech-to-text data and speech-to-text labels into training data, validation data, and test data; Perform data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; The ratio among the training set, the test set, and the validation set is 7:2:1;

[0083] Select an artificial intelligence model as the basic model; The artificial intelligence model includes an ASR model, etc.;

[0084] Train the basic model through the training set and adjust the learning rate and hyperparameters on the validation set to obtain a pre-trained model;

[0085] By validating the pre-trained model on the test set, a speech recognition model that takes input speech data and outputs speech-text data and speech-text labels is finally obtained.

[0086] The large language model in this embodiment is obtained by fine-tuning on government domain data, including:

[0087] Obtain a pre-trained large language model and government domain data; the government domain data includes several historical text data and their corresponding work order data, speech-text data and speech-text labels and their corresponding work order data; the work order data includes consultation time, work order location, duration, work order content, content level, and affiliated department, etc.;

[0088] Divide several historical text data and their corresponding work order data, speech-text data and speech-text labels and their corresponding work order data into training data, validation data, and test data; perform data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; the ratio among the training set, the test set, and the validation set is 7:2:1;

[0089] Train the pre-trained large language model with the training set and adjust the learning rate and hyperparameters on the validation set; the pre-trained large language model includes models such as ChatGLM;

[0090] By validating the pre-trained large language model on the test set, a large language model that takes input text data or speech-text data and their corresponding speech-text labels and outputs work order data is finally obtained.

[0091] In this embodiment, by constructing a speech recognition model, the demands of people in different regions are accurately recognized, and the text data and their corresponding dialect features are recognized from the speech. On this basis, work order data is obtained through the large language model; among them, the large language model is trained with government domain data, which can enhance the accurate recognition of the large language model for user demands and provide good data support for subsequent work order dispatching.

[0092] After constructing a dispatching objective function based on the work order data in this embodiment and solving it, the obtained dispatching suggestions include:

[0093] Obtain work order data; the work order data includes consultation time, work order location, duration, work order content, content level, affiliated department, and their corresponding confidence level ZD;

[0094] Generate several target quantization functions according to the duration, content level, affiliated department, and their corresponding confidence level;

[0095] Through the formula Construct the dispatch objective function FC; where ω1, ω2, and ω3 are dynamic weight coefficients, ω1 ∈ (α1, α2), ω2 ∈ (α3, α4), ω3 ∈ (α5, α6); the dynamic weight coefficients are generated according to the work order data; α1, α2, α3, α4, α5, and α6 are constants, and α1, α2, α3, α4, α5, and α6 ∈ (0, 1), and the specific values are set according to experience. The setting of the constants is to enable the dynamic weight coefficients to be adjusted within a certain range to avoid the situation where the objective function corresponding to a certain dynamic weight coefficient becomes ineffective; DJ is the maximization level quantization function, DD is the unit level; QT is the maximization deadline occupancy ratio quantization function, DT is the unit time ratio; Z is the maximization confidence quantization function, DZ is the unit confidence; BRL is the maximization department task completion rate, DL is the unit completion rate;

[0096] Generate dispatch suggestions according to the dispatch objective function.

[0097] In this embodiment, several objective quantization functions are generated according to the deadline duration, content level, and the department to which it belongs and its corresponding confidence, including:

[0098] Obtain the work order waiting dispatch time in real time, and obtain the consultation time, deadline duration QS, content level, the department to which it belongs and its corresponding confidence, and the department completion rate BWL;

[0099] Construct the maximization level quantization function ND through the formula ND = max{DJ i}; where DJ i represents the content level of the i-th work order data;

[0100] Obtain the waiting duration DS by taking the difference between the work order waiting dispatch time and the consultation time;

[0101] Construct the minimization deadline occupancy ratio quantization function QT through the formula ; where DS i represents the waiting duration of the i-th work order data; QS i represents the deadline duration of the i-th work order data;

[0102] Construct the maximization confidence quantization function Z through the formula Z = max{ZD ij}; where ZD ij represents the j-th confidence in the i-th work order data; its goal is to select the most suitable department to which it belongs;

[0103] Construct the maximization department task completion rate BRL through the formula BRL = max{BWL j}; where BWL j represents the department completion rate of the department corresponding to the j-th confidence.

[0104] In this embodiment, multiple target quantization functions are constructed and integrated into a dispatch target function to comprehensively evaluate the time urgency, level importance of the work order, and the confidence level of the department to which it belongs. At the same time, based on considering the department to which it belongs, the system comprehensively evaluates the task completion rate of this department, accurately screens out the work order data to be dispatched and its most suitable dispatch department, thereby greatly improving the intelligence level and dispatch efficiency of the work order dispatch system.

[0105] The dynamic weight coefficient in this embodiment is generated according to the work order data, including:

[0106] Obtain work order data; the work order data includes the deadline duration, content level, and the department to which it belongs and its corresponding confidence level;

[0107] Input the work order data into the weight estimation model to obtain the corresponding dynamic weight coefficient;

[0108] Among them, the weight estimation model is constructed through a machine learning model, including:

[0109] Obtain a number of historical work order data and their corresponding historical weight coefficients;

[0110] Divide a number of historical work order data and their corresponding historical weight coefficients into training data, validation data, and test data; perform data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; the ratio among the training set, the test set, and the validation set is 7:2:1;

[0111] Select a machine learning model as the basic model; the machine learning model includes a convolutional neural network model, etc.;

[0112] Train the basic model through the training set, and adjust the learning rate and hyperparameters on the validation set to obtain a pre-trained model;

[0113] Verify the pre-trained model on the test set, and finally obtain a weight estimation model that inputs work order data and outputs a number of dynamic weight coefficients.

[0114] Through the above steps in this embodiment, the dynamic weight coefficient is updated in real time, enabling the system to adaptively adjust the appropriate weight coefficient under different work order data, thereby improving the accuracy and efficiency of the work order dispatch system.

[0115] Generating dispatch suggestions according to the dispatch target function in this embodiment includes:

[0116] Obtain the dispatch target function;

[0117] The optimal solution of the dispatch objective function is obtained by passing production data and product comprehensive functions through a deep reinforcement learning model, and dispatch suggestions are generated based on this; the deep reinforcement learning model is a pre-trained model that can effectively obtain the optimal solution of the dispatch objective function and generate dispatch suggestions based on this.

[0118] Generating an alarm signal according to the processing duration in this embodiment includes:

[0119] Obtaining the processing duration and deadline duration of the work order data in real time;

[0120] Judging whether the processing duration is less than the deadline duration;

[0121] If yes, judge whether the processing duration is greater than D times the deadline duration; if yes, generate a warning signal that the processing is about to exceed the deadline; if no, do nothing;

[0122] If no, generate an alarm signal that the processing has exceeded the deadline; where D is a proportionality coefficient, D ∈ (0, 1), and the specific value is set according to experience. In this embodiment, D is set to 0.8.

[0123] Please refer to Figure 2 , another embodiment of the present application provides a method for dispatching government service hotline work orders based on a large language model, including:

[0124] S0: Obtain hotline data;

[0125] S1: Generate work order data according to the hotline data;

[0126] S2: After constructing a dispatch objective function based on the work order data, solve to obtain dispatch suggestions; dispatch the work order data according to the dispatch suggestions;

[0127] S3: Obtain the processing duration of the work order data in real time, and generate an alarm signal according to the processing duration;

[0128] S4: Make a prompt according to the alarm signal and contact the management personnel.

[0129] Some of the data in the above formula are calculated by removing the dimension and taking its numerical value. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; the preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0130] Working principle of this application: Obtain hotline data; Generate work order data based on the hotline data; Construct a dispatch objective function according to the work order data and then solve it to obtain dispatch suggestions; Dispatch the work order data according to the dispatch suggestions; Real-time obtain the processing duration of the work order data, generate an alarm signal according to the processing duration; Make a prompt according to the alarm signal and contact the management personnel, taking into account the dialect characteristics of each place, so that the work order dispatch system can accurately identify the demands of different groups of people. At the same time, construct a dispatch objective function, comprehensively consider multiple factors to enable the work order to be dispatched in a timely and effective manner; Improve the accuracy and efficiency of the work order dispatch system, avoid the problems that the existing technology often relies on manual judgment and experience, and at the same time lacks consideration of whether the department to be dispatched can process the work order within the effective time, resulting in low accuracy and efficiency of the work order dispatch system.

[0131] The above embodiments are only used to illustrate the technical method of this application and not to limit it. Although this application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of this application can be modified or equivalently replaced without departing from the spirit and scope of the technical method of this application.

Claims

1. A government hotline work order dispatching system based on a large language model, characterized in that, Including: A data collection module, a data analysis module, an early warning module, and a database; The data collection module: Obtains hotline data through data collection devices; The data analysis module: Generates work order data based on the hotline data; Constructs a dispatch objective function based on the work order data and then solves to obtain a dispatch suggestion; Dispatches the work order data according to the dispatch suggestion; Realtime obtains the processing duration of the work order data and generates an alarm signal according to the processing duration; The early warning module: Gives a prompt according to the alarm signal and contacts the management personnel; The database is used to store historical data required for training the model.

2. The government affairs hotline work order dispatching system based on the large language model according to claim 1, wherein The generating of work order data based on the hotline data includes: Obtaining hotline data; The hotline data includes text data and voice data; Judging whether the hotline data is voice data; If yes, inputs the voice data into a speech recognition model to obtain speech-to-text data and speech-to-text labels; The speech recognition model is constructed through an artificial intelligence model; If not, inputs the text data or the speech-to-text data and its corresponding speech-to-text labels into a large language model to obtain work order data; The large language model is obtained by fine-tuning on government domain data.

3. The government affairs hotline work order dispatching system based on the large language model according to claim 2, wherein, The construction of the speech recognition model through an artificial intelligence model includes: Obtaining a number of historical voice data and their corresponding historical speech-to-text data and speech-to-text labels; Dividing a number of historical voice data and their corresponding historical speech-to-text data and speech-to-text labels into training data, validation data, and test data; Performing data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; Selecting an artificial intelligence model as the basic model; Training the basic model through the training set and adjusting the learning rate and hyperparameters on the validation set to obtain a pre-trained model; Verifying the pre-trained model on the test set, and finally obtaining a speech recognition model that inputs voice data and outputs speech-to-text data and speech-to-text labels.

4. The government affairs hotline work order dispatching system based on the large language model according to claim 2, characterized in that, The obtaining of the large language model by fine-tuning on government domain data includes: Obtaining a pre-trained large language model and government domain data; The government domain data includes a number of historical text data and their corresponding work order data, speech-to-text data and speech-to-text labels and their corresponding work order data; Dividing a number of historical text data and their corresponding work order data, speech-to-text data and speech-to-text labels and their corresponding work order data into training data, validation data, and test data; Performing data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; Training the pre-trained large language model through the training set and adjusting the learning rate and hyperparameters on the validation set; Verifying the pre-trained large language model on the test set, and finally obtaining a large language model that inputs text data or speech-to-text data and their corresponding speech-to-text labels and outputs work order data.

5. The government affairs hotline work order dispatching system based on the large language model according to claim 1, characterized in that, The constructing of a dispatch objective function based on the work order data and then solving to obtain a dispatch suggestion includes: Obtaining work order data; The work order data includes the consultation time, work order location, deadline duration, work order content, content level, and the affiliated department and its corresponding confidence level ZD; Generating a number of target quantization functions according to the deadline duration, content level, and the affiliated department and its corresponding confidence level; Through the formula Construct the distribution objective function FC; where ω1, ω2, and ω3 are dynamic weight coefficients, ω1 ∈ (α1, α2), ω2 ∈ (α3, α4), ω3 ∈ (α5, α6); the dynamic weight coefficients are generated according to the work order data; α1, α2, α3, α4, α5, and α6 are constants, and α1, α2, α3, α4, α5, and α6 ∈ (0, 1); DJ is the maximized level quantization function, DD is the unit level; QT is the maximized deadline occupancy ratio quantization function, DT is the unit time ratio; Z is the maximized confidence quantization function, DZ is the unit confidence; BRL is the maximized department task completion rate, DL is the unit completion rate; Generate dispatch suggestions according to the dispatch objective function.

6. The government affairs hotline work order dispatching system based on a large language model according to claim 5, characterized in that, Generate a number of target quantization functions based on the deadline duration, content level, department and its corresponding confidence level, including: Obtain the time when the work order is to be dispatched in real time, and obtain the consultation time, deadline duration QS, content level, department and its corresponding confidence level, and department completion rate BWL; Construct the maximum-level quantization function ND through the formula ND = max{DJ i}; where DJ i represents the content level of the i-th work order data. Obtain the waiting duration DS by subtracting the consultation time from the time when the work order is to be dispatched; Through the formula Construct the minimization deadline occupancy ratio quantization function QT; where DS i Represents the waiting duration of the i-th work order data; QS i Represents the deadline duration of the i-th work order data; Construct the maximum confidence quantization function Z through the formula Z = max{ZD ij}; where ZD ij represents the j-th confidence in the i-th work order data. Construct the maximum department task completion rate BRL through the formula BRL = max{BWL j}; where BWL j represents the department completion rate of the department corresponding to the j-th confidence level.

7. The government affairs hotline work order dispatching system based on a large language model according to claim 5, characterized in that, The dynamic weight coefficient is generated according to the work order data, including: Obtain the work order data; the work order data includes the deadline duration, content level, department and its corresponding confidence level; Input the work order data into the weight prediction model to obtain the corresponding dynamic weight coefficient; Among them, the weight prediction model is constructed by a machine learning model, including: Obtain a number of historical work order data and their corresponding historical weight coefficients; Divide a number of historical work order data and their corresponding historical weight coefficients into training data, validation data, and test data; perform data preprocessing on the training data, validation data, and test data to obtain a training set, a validation set, and a test set; Select a machine learning model as the basic model; Train the basic model with the training set, and adjust the learning rate and hyperparameters on the validation set to obtain a pre-trained model; Verify the pre-trained model on the test set, and finally obtain the input work order data, and the output is a weight prediction model of a number of dynamic weight coefficients.

8. The government affairs hotline work order dispatching system based on the large language model according to claim 5, characterized in that, The generation of dispatch suggestions according to the dispatch objective function includes: Obtain the dispatch objective function; Obtain the optimal solution of the dispatch objective function by passing the production data and the product comprehensive function through the deep reinforcement learning model, and generate dispatch suggestions based on this.

9. The government affairs hotline work order dispatching system based on a large language model according to claim 1, wherein The generation of an alarm signal according to the processing duration includes: Obtain the processing duration and deadline duration of the work order data in real time; Judge whether the processing duration is less than the deadline duration; If yes, judge whether the processing duration is greater than D times the deadline duration; if yes, generate a warning signal that the processing is about to exceed the deadline; if not, do nothing; If not, generate an alarm signal that the processing has exceeded the deadline; where D is a proportionality coefficient, D∈(0, 1).

10. A method for dispatching government affairs hotline work orders based on a large language model, applied to the government affairs hotline work order dispatching system based on the large language model according to any one of claims 1-9, characterized in that, Including: S0: Obtain hotline data; S1: Generate work order data according to the hotline data; S2: Construct a dispatch objective function according to the work order data and solve it to obtain dispatch suggestions; dispatch the work order data according to the dispatch suggestions; S3: Obtain the processing duration of the work order data in real time, and generate an alarm signal according to the processing duration; S4: Make a prompt according to the alarm signal and contact the management personnel.

Images