A Mail Delivery Method and System Based on a Large Model

Abstract

This invention provides a mail delivery method and system based on a large-scale model. The method first configures and stores information from multiple departments in a preset database to enable efficient invocation and processing of subsequent workflows. When a citizen receives a mail to be delivered and determines it is not a text mail, it is converted into a text mail for easier data analysis. Then, based on the entered department information, the target department corresponding to the text mail is extracted from the preset database, and the corresponding mail delivery operation is executed. Intelligent matching through the large-scale model enables accurate mail delivery, improving the efficiency of mail processing. Finally, historical similar mail associated with the text mail to be delivered is queried from the preset database, providing citizens with historical reference information. This process achieves efficient and automatic mail delivery for citizens, avoiding the problem of misdelivery due to subjective judgment in traditional manual delivery methods, and significantly improving the accuracy of mail delivery.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and system for mail delivery based on a large model. Background Technology

[0003] Currently, traditional methods of delivering citizen mail still rely mainly on manual labor, requiring delivery personnel to manually analyze and manage mail delivery based on fixed rules and keyword matching.

[0004] However, when faced with a massive volume of citizen mail, the traditional manual delivery method is slow to process mail and is easily affected by subjective judgment, thus reducing the accuracy of mail delivery. Summary of the Invention

[0005] This invention provides a mail delivery method and system based on a large model, which can improve the accuracy of mail delivery.

[0006] In a first aspect, embodiments of the present invention provide a letter dispatch method based on a large model, the method comprising: At least one department information is pre-configured and stored in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Upon receiving a pending email from a current citizen, determine whether the pending email is a text email. When it is determined that the message to be dispatched is not the text message, the message to be dispatched is converted into a text message to be dispatched. Based on the information of each department, the target department corresponding to the text letter to be assigned is determined from the preset database and the text letter to be assigned is assigned to the target department; The large model is used to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen. Each historical similar letter includes: a description of the historical problem, the historical processing department, the historical processing result, and citizen feedback evaluation.

[0007] Preferably, After determining whether the pending email is a text message upon receiving a current citizen's pending email, the process further includes: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

[0008] Preferably, When it is determined that the message to be dispatched is not a text message, converting the message to be dispatched into a text message to be dispatched includes: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

[0009] Preferably, The process involves using the large model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and sending it to the current citizen. Each historical similar letter includes: a description of the historical issue, the department responsible for handling the historical issue, the result of the historical handling, and citizen feedback and evaluation. Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

[0010] Preferably, After querying the preset database using the large model for at least one historically similar letter associated with the text letter to be dispatched and sending it to the current citizen, the process further includes: Upon receiving a request to add a new department from an external source, the request is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

[0011] Secondly, embodiments of the present invention provide a mail delivery system based on a large model, the system comprising: Department Information Input Module: Used to pre-configure at least one department information and store it in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Letter recognition module: used to determine whether a letter to be assigned is a text letter when a citizen receives a letter to be assigned. Non-text message processing module: used to convert the message to be dispatched into a text message when the message recognition module determines that the message to be dispatched is not a text message; The letter dispatch module is used to determine the target department corresponding to the text letter to be dispatched converted by the non-text letter processing module from the preset database based on the department information entered by the department information entry module, and dispatch the text letter to be dispatched to the target department. Similar letter recommendation module: used to use the big model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen, wherein each historical similar letter includes: the historical problem description, the historical processing department, the historical processing result, and the citizen's feedback evaluation.

[0012] Preferably, Following the letter recognition module, a text letter processing module is further included; The text message processing module is used to perform: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

[0013] Preferably, The non-text message processing module is also used to perform: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

[0014] Preferably, The similar letter recommendation module is also used to perform: Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

[0015] Preferably, Following the similar letter push module, the system further includes: a department information update module; The department information update module is used to perform: Upon receiving a request to add a new department from an external source, the request is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

[0016] This invention provides a method and system for mail delivery based on a large-scale model. The method first configures and stores departmental information, including department name, departmental description, departmental responsibilities, and historical processing cases, in a preset database to enable efficient invocation and processing of subsequent procedures. Next, upon receiving a mail to be delivered from a citizen and determining that it is not a text mail, it is converted into a text mail for easier data analysis. Then, based on pre-entered departmental responsibility information, the target department corresponding to the text mail is extracted from the preset database, and the corresponding mail delivery operation is executed. Intelligent matching through the large-scale model enables accurate mail delivery, further improving mail processing efficiency. Finally, the large-scale model queries the preset database for multiple historically similar mails associated with the text mail and sends them to the citizen, increasing the transparency of the delivery process and providing historical reference information. This allows citizens to form reasonable expectations regarding the resolution process and expected results of their mail, enhancing their trust and satisfaction with government services. The above process, through departmental information entry, non-text letter processing, intelligent matching of large models, and intelligent letter distribution based on recommendations of similar historical letters, solves the problems of inefficient resource allocation and inaccurate response in the current government's handling of various issues and opinions in citizens' letters. It can achieve efficient and automatic distribution of citizens' letters, while also avoiding the problem of misdelivery caused by subjective judgment in the traditional manual letter distribution method, which can significantly improve the accuracy of letter distribution. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a letter delivery method based on a large model provided in an embodiment of the present invention; Figure 2 This is a flowchart of another letter delivery method based on a large model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a mail delivery system based on a large model provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of another mail delivery system based on a large model provided in an embodiment of the present invention; Figure 5This is a schematic diagram of another mail delivery system based on a large model provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, this embodiment of the invention provides a letter delivery method based on a large model, which may include the following steps: Step 101: Pre-configure at least one department information and store it in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Step 102: Upon receiving a pending letter from a current citizen, determine whether the pending letter is a text message. Step 103: If it is determined that the message to be dispatched is not a text message, convert the message to be dispatched into a text message to be dispatched. Step 104: Based on the information of each department, determine the target department corresponding to the text letter to be assigned from the preset database and assign the text letter to be assigned to the target department; Step 105: Use the large model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen. Each historical similar letter includes: historical problem description, historical processing department, historical processing result, and citizen feedback evaluation.

[0021] This invention provides a mail delivery method based on a large-scale model. The method first configures and stores departmental information, including department name, departmental description, departmental responsibilities, and historical processing cases, in a preset database to enable efficient invocation and processing in subsequent workflows. Next, upon receiving a mail to be delivered from a citizen and determining that it is not a text mail, it is converted into a text mail for easier data analysis. Then, based on pre-entered departmental responsibility information, the target department corresponding to the text mail is extracted from the preset database, and the corresponding mail delivery operation is executed. Intelligent matching through the large-scale model enables accurate mail delivery, further improving mail processing efficiency. Finally, the large-scale model queries the preset database for multiple historically similar mails associated with the text mail and sends them to the citizen, increasing the transparency of the delivery process and providing historical reference information. This allows citizens to form reasonable expectations regarding the resolution process and expected results of their mail, enhancing their trust and satisfaction with government services. The above process, through departmental information entry, non-text letter processing, intelligent matching of large models, and intelligent letter distribution based on recommendations of similar historical letters, solves the problems of inefficient resource allocation and inaccurate response in the current government's handling of various issues and opinions in citizens' letters. It can achieve efficient and automatic distribution of citizens' letters, while also avoiding the problem of misdelivery caused by subjective judgment in the traditional manual letter distribution method, which can significantly improve the accuracy of letter distribution.

[0022] In order to process text messages, in one embodiment of the present invention, after step 102 in the above embodiment, the following steps are further included: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

[0023] In this embodiment of the invention, to process text letters, upon determining that the letter to be dispatched is a text letter, a unified text preprocessing operation (e.g., typo identification and correction) can be performed on the letter to be dispatched using the fast processing and natural language understanding capabilities of a large model (e.g., an LLM language model). This ensures the standardization and consistency of the data, facilitating efficient subsequent retrieval and processing. While ensuring the accuracy of the text information, this also shortens the letter processing time to meet the real-time dispatch requirements of a large number of letters. Then, text enhancement operations can be performed on the preprocessed letter to be dispatched using the language model to generate the text letter to be dispatched, through text polishing and text... Optimization and other methods are used to enhance the richness and adaptability of text information, ensuring its integrity. Finally, a first letter index containing letter keywords and associated personnel can be created based on the text letters to be dispatched. This allows for subsequent deep semantic analysis and contextual understanding based on the semantic understanding and generalization capabilities of the language big data model, accurately capturing the true intentions hidden in citizens' letters, and constructing an accurate mapping relationship between contextual information and departmental information. By accurately matching the first letter index, including letter keywords and associated personnel, with the corresponding target department's responsibilities, a reliable data foundation is laid for subsequent data processing, further improving the efficiency of text letter processing.

[0024] In order to process non-text messages, in one embodiment of the present invention, step 103 in the above embodiment may specifically include the following steps: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

[0025] In this embodiment of the invention, since citizen letters often contain non-text information such as images and videos, traditional letter processing methods suffer from the difficulty of processing multimodal information, making it difficult to effectively extract key information from the letters and incorporate it into the delivery decision. Therefore, when it is determined that the letter to be delivered is not a text letter, the non-text information in the letter to be delivered is first parsed to effectively process images, videos, and other non-text information, laying a reliable data foundation for subsequent text conversion operations. Then, a large model (e.g., an MLLM multimodal large model) is used to perform multimodal processing operations on the non-text information in the letter to be delivered, converting it into a unified text format (e.g., using an Illava multimodal large model to convert image content into text descriptions), providing a solid data foundation for subsequent intelligent delivery. Next, the text description is associated and bound with the letter to be delivered to generate a text letter to be delivered containing the text description, enabling data traceability and retrieving the corresponding text information through letter retrieval. Finally, a second letter index containing the text description is created based on the text letter to be delivered, enabling subsequent fast retrieval of non-text letters through the second letter index. The above process uses deep semantic understanding and multimodal fusion technologies of large models and integrates multimodal data for comprehensive judgment to achieve efficient and accurate classification and intelligent distribution of citizens' letters. This can improve the comprehensiveness and accuracy of letter distribution. At the same time, the preprocessed text and the text description obtained from other modal conversions will also be input into the large model to perform subsequent department matching operations.

[0026] In order to obtain historically similar letters, in one embodiment of the present invention, step 105 in the above embodiment may specifically include the following steps: Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

[0027] In this embodiment of the invention, to obtain historical similar letters, after a citizen submits a letter to be assigned, the large model can extract key information from the letter (such as the core content of the problem, the objects involved, etc.) to retrieve corresponding historical similar letters in a preset database. Based on this, the text letter to be assigned and each historical letter in the preset database can be converted into a text letter vector to be assigned and a historical letter vector, respectively. Then, the letter similarity between the text letter vector to be assigned and each historical letter vector is calculated using the first formula mentioned above, and multiple letter similarities that meet the threshold condition are determined based on a preset similarity threshold. Finally, based on the set first and second letter indexes and the multiple letter similarities calculated above, a batch of historical similar letters with high similarity are selected from the preset database. These historical similar letters contain detailed information such as historical problem descriptions, historical processing departments, historical processing results, and citizen feedback and evaluations. This information is then organized and recommended to the current citizen, allowing the citizen to intuitively understand the processing process of historical cases similar to their own problems, thereby having a reasonable expectation of the problem-solving process and expected results, and further improving the citizen's satisfaction with government services.

[0028] To improve the adaptability and flexibility of the system, in one embodiment of the present invention, after step 105, the above embodiment further includes the following steps: Upon receiving a request to add a new department from an external source, the request is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

[0029] In this embodiment of the invention, since departmental information is complex, diverse, and constantly changing, and the adjustment of functions and the addition of new departments are common occurrences in government departments, when a request to add a new department is received from an external source, the request must first be parsed to obtain key information, laying the data foundation for subsequent processing of the new department process. Then, based on preset configuration rules, multiple target field parameters (e.g., new department name parameter, new department description parameter, new department scope of responsibilities parameter) corresponding to multiple target fields in the new department request are extracted, laying the data foundation for subsequent departmental information entry. Next, data standardization processing is performed on the multiple target field parameters to generate multiple standard target field parameters, ensuring data standardization and consistency. Finally, the multiple standard target field parameters and their corresponding multiple target fields are associated and stored in the departmental information configuration table in the preset database. This enables dynamic updates of new departments and facilitates subsequent intelligent departmental matching. Government departments can add new departments or modify existing departmental information at any time according to actual conditions without retraining the classification model, reducing system maintenance costs and significantly enhancing the system's adaptability and flexibility.

[0030] For example, when a department adds a new function or a special task force is temporarily established to deal with public emergencies, the relevant department managers only need to modify the corresponding parameters in the corresponding target field of the department information configuration table in the system to update the scope of department responsibilities and automatically adjust the subsequent mail dispatch logic to ensure that citizens' relevant issues can be accurately assigned to the correct department for processing.

[0031] like Figure 2 As shown, in order to more clearly illustrate the technical solution and advantages of the present invention, the following provides a detailed description of the letter delivery method based on a large model, which may include the following steps: Step 201: Pre-configure at least one department information and store it in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Step 202: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched. The text preprocessing includes: typo identification and correction. Step 203: Perform text enhancement operations on the preprocessed emails to be dispatched to generate the emails to be dispatched. The text enhancement operations include: text polishing and text optimization. Step 204: Create a first letter index containing letter keywords and associated personnel based on the text letters to be dispatched, and then proceed to step 209; Step 205: When it is determined that the letter to be dispatched is not a text letter, the large model is used to parse the non-text information in the letter to be dispatched. The non-text information includes: images and videos. Step 206: Perform multimodal processing on the non-text information in the emails to be dispatched, so as to convert the non-text information in the emails to be dispatched into text descriptions; Step 207: Associate and bind the text description with the email to be dispatched to generate an email to be dispatched containing the text description; Step 208: Create a second letter index containing text descriptions based on the text letters to be dispatched; Step 209: Based on the information of each department, determine the target department corresponding to the text letter to be assigned from the preset database and assign the text letter to be assigned to the target department; Step 210: Convert the text emails to be dispatched into a text email vector; Step 211: Synchronously convert each historical letter in the preset database into a historical letter vector; Step 212: Based on the text message vector to be assigned and each historical message vector, calculate the message similarity between the text message vector to be assigned and each historical message vector using the first formula; Specifically, the first formula is: ;in, For letter similarity, A is the vector of the text letter to be assigned, and B is the vector of each historical letter; Step 213: Based on the preset similarity threshold, the first letter index, and the second letter index, filter out at least one historical similar letter from the preset database whose letter similarity is greater than the preset similarity threshold; Step 214: Send at least one historically similar letter to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the department that handled the historical issue, the result of the historical handling, and citizen feedback and evaluation; Step 215: Upon receiving a request to add a new department from an external source, parse the request. Step 216: Obtain the parsed new department request, and extract at least one target field parameter corresponding to at least one target field in the new department request based on the preset configuration rules. The at least one target field parameter includes: new department name parameter, new department description parameter, and new department responsibility scope parameter. Step 217: Perform data standardization on at least one target field parameter to generate at least one standard target field parameter; Step 218: Associate at least one standard target field parameter with at least one corresponding target field and store them in the department information configuration table in the preset database.

[0032] like Figure 3 As shown, this embodiment of the invention provides a mail delivery system based on a large model, the system comprising: Department information entry module 301: used to pre-configure at least one department information and store it in a preset database, wherein each department information includes at least: department name, department introduction, department responsibilities, and historical processing cases; The letter identification module 302 is used to determine whether the letter to be assigned is a text letter when it receives a letter to be assigned from a current citizen. Non-text letter processing module 303: When the letter recognition module 302 determines that the letter to be dispatched is not a text letter, it converts the letter to be dispatched into a text letter to be dispatched. The letter dispatching module 304 is used to determine the target department corresponding to the text letter to be dispatched converted by the non-text letter processing module 303 from the preset database based on the department information entered by the department information entry module 301, and dispatch the text letter to be dispatched to the target department. Similar letter recommendation module 305: used to use the big model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen, wherein each historical similar letter includes: the historical problem description, the historical processing department, the historical processing result, and the citizen feedback evaluation.

[0033] based on Figure 3 The letter delivery system based on a large model shown is as follows: Figure 4 The system further includes, after the letter recognition module 302, a text letter processing module 306. The text message processing module 306 is used to perform: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

[0034] like Figure 3 As shown, the non-text message processing module 303 is also used to perform: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

[0035] like Figure 3 As shown, the similar letter recommendation module 305 is also used to perform: Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

[0036] based on Figure 4 The letter delivery system based on a large model shown is as follows: Figure 5 As shown, after the similar letter push module 305, a department information update module 307 is further included; The department information update module 307 is used to perform: Upon receiving a request to add a new department from an external source, the request to add a new department is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

[0037] It is understood that the structures illustrated in the embodiments of the present invention do not constitute a specific limitation on the large-model-based mail delivery system. In other embodiments of the present invention, the large-model-based mail delivery system may include more or fewer components than illustrated, or combine some components, split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0038] The information interaction and execution process between the various units in the above-mentioned device are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description of the method embodiment of the present invention, and will not be repeated here.

[0039] This invention also provides a mail delivery system based on a large model, comprising: at least one memory and at least one processor; At least one memory for storing machine-readable programs; At least one processor is configured to invoke a machine-readable program to execute the large-model-based letter delivery method according to any embodiment of the present invention.

[0040] This invention also provides a computer-readable medium storing computer instructions, which, when executed by a processor, cause the processor to perform the mail delivery method based on a large model according to any embodiment of this invention.

[0041] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0042] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0043] Storage media embodiments for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0044] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0045] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion unit connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion unit execute some and all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0046] The various embodiments of the present invention have at least the following beneficial effects: 1. This invention provides a mail delivery method based on a large model. The method first configures and stores departmental information, including department name, departmental description, departmental responsibilities, and historical processing cases, in a preset database to enable efficient invocation and processing in subsequent workflows. Next, upon receiving a mail to be delivered from a citizen and determining that it is not a text mail, it is converted into a text mail for easier data analysis. Then, based on pre-entered departmental responsibility information, the target department corresponding to the text mail is extracted from the preset database, and the corresponding mail delivery operation is executed. Intelligent matching through the large model enables accurate mail delivery, further improving mail processing efficiency. Finally, the large model queries the preset database for multiple historically similar mails associated with the text mail and sends them to the citizen, increasing the transparency of the delivery process and providing historical reference information. This allows citizens to form reasonable expectations regarding the resolution process and expected results of their mail, enhancing their trust and satisfaction with government services. The above process, through departmental information entry, non-text letter processing, large-scale intelligent matching, and intelligent letter distribution based on recommendations of similar historical letters, solves the problems of inefficient resource allocation and inaccurate response in the current government's handling of various issues and opinions in citizens' letters. It can achieve efficient and automatic distribution of citizens' letters, while also avoiding the problem of misdelivery caused by subjective judgment in the traditional manual letter distribution method, which can significantly improve the accuracy of letter distribution. 2. In this embodiment of the invention, in order to process text letters, when it is determined that the letter to be dispatched is a text letter, the large-scale model's fast processing and natural language understanding capabilities can be used to perform a unified text preprocessing operation on the letter to be dispatched to ensure the standardization and consistency of the data, so as to facilitate efficient subsequent calling and processing. While ensuring the accuracy of text information, it can also shorten the letter processing time to meet the real-time dispatch requirements of a large number of letters. Then, relying on the language large-scale model, text enhancement operations can be performed on the preprocessed letter to be dispatched to generate the text letter to be dispatched. This can be achieved by improving the richness and adaptability of text information through text polishing, text optimization, etc., to ensure the integrity of text information. Finally, a first letter index containing letter keywords and associated personnel can be created based on the text letter to be dispatched. This allows for subsequent deep semantic analysis and contextual understanding based on the semantic understanding and generalization capabilities of the language large-scale model, accurately capturing the true intentions hidden in citizens' letters, and constructing an accurate mapping relationship between contextual information and departmental information. By accurately matching the first letter index, such as letter keywords and associated personnel, with the corresponding target department's responsibilities, a reliable data foundation is laid for subsequent data processing, and the efficiency of text letter processing is further improved. 3. In this embodiment of the invention, since citizen letters often contain non-text information such as images and videos, traditional letter processing methods suffer from the difficulty of processing multimodal information, making it difficult to effectively extract key information from the letters and integrate it into the delivery decision. Therefore, when it is determined that the letter to be delivered is not a text letter, the non-text information in the letter to be delivered is first parsed to effectively process non-text information such as images and videos, laying a reliable data foundation for subsequent text conversion operations. Then, a large model is used to perform multimodal processing operations on the non-text information in the letter to be delivered, converting it into a unified text format, providing a solid data foundation for subsequent intelligent delivery. Next, the text description is associated and bound with the letter to be delivered to generate a text letter to be delivered containing the text description, enabling data traceability and retrieving the corresponding text information through letter retrieval. Finally, a second letter index containing the text description is created based on the text letter to be delivered, enabling subsequent fast retrieval of non-text letters through the second letter index. The above process uses deep semantic understanding and multimodal fusion technologies of large models and integrates multimodal data for comprehensive judgment to achieve efficient and accurate classification and intelligent distribution of citizens' letters. This can improve the comprehensiveness and accuracy of letter distribution. At the same time, the preprocessed text and the text description obtained from other modal conversions will also be input into the large model to perform subsequent department matching operations.

[0047] It should be noted that not all steps and modules in the above processes and system structure diagrams are mandatory; some steps or modules can be omitted as needed. The execution order of each step is not fixed and can be adjusted as required. The system structure described in the above embodiments can be a physical structure or a logical structure. That is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or they may be jointly implemented by certain components in multiple independent devices.

[0048] In the above embodiments, the hardware units can be implemented mechanically or electrically. For example, a hardware unit may include permanent dedicated circuitry or logic (such as a dedicated processor, FPGA, or ASIC) to perform the corresponding operation. The hardware unit may also include programmable logic or circuitry (such as a general-purpose processor or other programmable processor), which can be temporarily configured by software to perform the corresponding operation. The specific implementation method (mechanical, dedicated permanent circuitry, or temporarily configured circuitry) can be determined based on cost and time considerations.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A letter delivery method based on a large model, characterized in that, The method includes: At least one department information is pre-configured and stored in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Upon receiving a pending email from a current citizen, determine whether the pending email is a text email. When it is determined that the message to be dispatched is not the text message, the message to be dispatched is converted into a text message to be dispatched. Based on the information of each department, the target department corresponding to the text letter to be assigned is determined from the preset database and the text letter to be assigned is assigned to the target department; The large model is used to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen. Each historical similar letter includes: a description of the historical problem, the historical processing department, the historical processing result, and citizen feedback evaluation.

2. The method according to claim 1, characterized in that, After determining whether the pending email is a text message upon receiving a current citizen's pending email, the process further includes: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

3. The method according to claim 1, characterized in that, When it is determined that the message to be dispatched is not a text message, converting the message to be dispatched into a text message to be dispatched includes: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

4. The method according to claim 1, characterized in that, The process involves using the large model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and sending it to the current citizen. Each historical similar letter includes: a description of the historical issue, the department responsible for handling the historical issue, the result of the historical handling, and citizen feedback and evaluation. Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

5. The method according to any one of claims 1-4, characterized in that, After querying the preset database using the large model for at least one historically similar letter associated with the text letter to be dispatched and sending it to the current citizen, the process further includes: Upon receiving a request to add a new department from an external source, the request is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

6. A mail delivery system based on a large model, characterized in that, The system includes: Department Information Input Module: Used to pre-configure at least one department information and store it in a preset database, wherein each department information includes at least: department name, department description, department responsibilities, and historical processing cases; Letter recognition module: used to determine whether a letter to be assigned is a text letter when a citizen receives a letter to be assigned. Non-text message processing module: used to convert the message to be dispatched into a text message when the message recognition module determines that the message to be dispatched is not a text message; The letter dispatch module is used to determine the target department corresponding to the text letter to be dispatched converted by the non-text letter processing module from the preset database based on the department information entered by the department information entry module, and dispatch the text letter to be dispatched to the target department. Similar letter recommendation module: used to use the big model to query at least one historical similar letter associated with the text letter to be dispatched from the preset database and send it to the current citizen, wherein each historical similar letter includes: the historical problem description, the historical processing department, the historical processing result, and the citizen's feedback evaluation.

7. The system according to claim 6, characterized in that, Following the letter recognition module, a text letter processing module is further included; The text message processing module is used to perform: When it is determined that the letter to be dispatched is a text letter, the large model is used to perform text preprocessing on the letter to be dispatched, wherein the text preprocessing includes: typo identification and correction; A text enhancement operation is performed on the pre-processed email to be dispatched to generate the email to be dispatched. The text enhancement operation includes: text polishing and text optimization. Based on the text message to be assigned, a first letter index containing letter keywords and associated personnel is created, and the process of determining the target department corresponding to the text message to be assigned from the preset database based on the department information is executed, and the text message to be assigned is assigned to the target department.

8. The system according to claim 6, characterized in that, The non-text message processing module is also used to perform: When it is determined that the letter to be dispatched is not the text letter, the large model is used to parse the non-text information in the letter to be dispatched, wherein the non-text information includes: images and videos; Multimodal processing is performed on the non-text information in the email to be dispatched to convert the non-text information in the email to be dispatched into a text description; The text description and the letter to be dispatched are associated and bound together to generate the text letter to be dispatched containing the text description; A second letter index containing the text description is created based on the text letter to be dispatched.

9. The system according to claim 6, characterized in that, The similar letter recommendation module is also used to perform: Convert the text message to be dispatched into a text message vector; Simultaneously convert each historical letter in the preset database into a historical letter vector; Based on the text message vector to be assigned and each of the historical message vectors, the message similarity between the text message vector to be assigned and each of the historical message vectors is calculated using the first formula. The first formula is: ; Among them, the The similarity of the letters is defined as follows: A is the vector of the text letter to be assigned, and B is the vector of each historical letter. Based on a preset similarity threshold, the first letter index, and the second letter index, at least one historical similar letter corresponding to a letter with a similarity greater than the preset similarity threshold is selected from the preset database. The at least one historically similar letter is sent to the current citizen, wherein each historically similar letter includes: a description of the historical issue, the historical processing department, the historical processing result, and the citizen's feedback and evaluation.

10. The system according to any one of claims 6-9, characterized in that, Following the similar letter push module, the system further includes: a department information update module; The department information update module is used to perform: Upon receiving a request to add a new department from an external source, the request is parsed. Obtain the parsed request for adding a department, and extract at least one target field parameter corresponding to at least one target field in the request for adding a department based on preset configuration rules. The at least one target field parameter includes: the name of the new department, the description of the new department, and the scope of responsibilities of the new department. Perform data standardization processing on the at least one target field parameter to generate at least one standard target field parameter; The at least one standard target field parameter and the corresponding at least one target field are associated and stored in the department information configuration table of the preset database.

Images