An education network user personalized mail processing method and system based on a large model

Abstract

The application discloses an education network user personalized mail processing method and system based on a large model, and belongs to the technical field of large models. The method is as follows: a user's social relationship database is constructed according to paper information, communication relationships and user historical operations in the last period; based on a currently received mail, a plurality of prompt words are extracted from the social relationship database; a preset large model is used to predict the user's mail receiving demand based on the prompt words, the currently received mail is analyzed according to the mail receiving demand, and the academic relevance degree and mail receiving suggestion of the currently received mail are output; and based on the academic relevance degree and mail receiving suggestion, the currently received mail is delivered to a corresponding mailbox directory using a preset mail screening condition. Therefore, by implementing the application, the problem of a high misjudgment rate of junk mail in the education network user scenario in the prior art can be solved.

Description

Technical Field

[0001] This application belongs to the field of large model technology, specifically relating to a method and system for personalized email processing for users of education networks based on large models. Background Technology

[0002] Unlike regular business users, users on educational networks often publish academic papers. Therefore, these papers can be used to obtain users' email addresses and send emails to them. As a result, educational network users frequently receive emails from unknown senders. Identifying spam from the large volume of emails received and promptly filtering out legitimate emails that are useful to users to improve their user experience is a pressing issue that needs to be addressed.

[0003] Current email identification and processing methods primarily rely on integrating multi-dimensional dynamic data such as recent user behavior, email content, and social relationships, along with dynamic judgment thresholds, to determine the importance of a current email to the user and whether it meets their current needs, intelligently identifying important emails from spam. However, in educational network scenarios, this method is prone to misjudging users' social relationships when sending mass emails, leading to many legitimate emails being deemed useless. Furthermore, it fails to consider the sending domains of some well-known research institutions (such as overseas educational network email domains, overseas academy of sciences email domains, and email domains of famous overseas research institutions), causing emails from these institutions to also be identified as spam. Summary of the Invention

[0004] This application proposes a personalized email processing method and system for education network users based on a large model, which can solve the problem of high false positive rate of spam in the education network user scenario in the prior art.

[0005] The first aspect of this application provides a method for processing personalized emails for users on educational networks based on a large model, the method comprising: Based on paper information, communication relationships, and user history operations from the previous period, a user social relationship database is constructed. Based on the currently received email, several prompt words are extracted from the social relationship database; wherein, the prompt words contain the sender's and user's historical academic information; Using a pre-defined large model, the system predicts the user's email receiving needs based on the prompt words, analyzes the currently received emails according to the email receiving needs, and outputs the academic relevance of the currently received emails and email receiving suggestions. Based on the academic relevance and email receiving suggestions, the currently received emails are delivered to the corresponding mailbox directory using preset email filtering conditions.

[0006] To accurately identify academic emails, the above solution constructs a social relationship database that accurately represents a user's recent preferences, social relationships, and relevant academic information, based on the paper information provided in the email, the user's communication relationships, and recent user actions. This database can determine a user's specific preferences for certain emails. Tip words related to the currently received emails are extracted from the social relationship database to assist the large-scale model in determining whether the email content meets the user's needs. Furthermore, the semantic understanding capabilities of the large-scale model are used to detect whether the currently received emails contain spam, enabling more accurate identification of academic emails with potential value to the user. This provides the academic relevance of the currently received emails and email delivery suggestions, thereby automatically delivering emails to the appropriate directories. This significantly improves the email processing efficiency for users on the education network, reduces the time cost of email filtering, and ensures that important academic information is not overlooked.

[0007] In one possible implementation of the first aspect, a user's social relationship database is constructed based on paper information, communication relationships, and the user's historical operations from the previous period, specifically as follows: Obtain the paper information corresponding to the existing email addresses from the existing public database, and construct a first database based on the mapping relationship between the existing email addresses and the paper information; Based on the user's historical operations, a second database is constructed according to the user's email report feedback information, a third database is constructed according to the user's email deletion information, and a fourth database is constructed according to the user's email movement information; wherein, the email movement information refers to moving emails from the spam folder to the inbox. Based on the email addresses of historical senders and users, a communication relationship between users and historical senders is constructed, and a fifth database is constructed based on the communication relationship. The first, second, third, fourth, and fifth databases are integrated to construct the social relationship database.

[0008] The above solution integrates data from multiple dimensions to construct a comprehensive and accurate database of user social and preference profiles. In building this database, publicly available academic information is incorporated, enabling the system to understand the sender's academic identity and field of expertise, which is fundamental to judging the academic value of emails. Simultaneously, based on recent user activity history, a database accurately reflecting recent user preferences is constructed, providing a genuine basis for personalized email processing. Furthermore, by recording the user's social relationships through communication with historical senders, a basis is provided for filtering spam emails from strangers.

[0009] One possible implementation of the first aspect also includes: By using the first database, paper information related to the email address currently receiving the email can be obtained, thus revealing the sender's published papers. The user's preferred email type is obtained through the second, third, and fourth databases.

[0010] In one possible implementation of the first aspect, based on the currently received email, several prompt words are extracted from the social relationship database, specifically as follows: Extract sender information, email title, and email body from currently received emails; Based on sender and user information, extract the titles of papers published by the sender and user from the social relationship database; Extract historical sender information from the social relationship database, as well as the email titles reported by the user in the previous period, the email titles of emails deleted without being read, and the email titles of emails moved from the spam folder to the inbox; The title of the paper, the email title, the email body, the sender history, and all the titles of the paper are used as the prompt words.

[0011] The above solution extracts highly relevant structured information from a social relationship database based on the information provided in the currently received emails. By extracting the titles of papers published by the sender and user, the large-scale model can establish academic connections and determine whether the email content is related to the shared research fields of both parties. By extracting historical email titles, the large-scale model can understand user behavior patterns. These prompts effectively guide the analysis direction of the large-scale model, making its generated academic relevance and email receiving suggestions more in line with the current needs of users on the education network.

[0012] In one possible implementation of the first aspect, the user's email receiving needs are predicted based on the prompt words, and the currently received emails are analyzed according to the email receiving needs to output the academic relevance of the currently received emails and email receiving suggestions, specifically: Based on the prompt words, detect whether the currently received email is related to the journal, and obtain the academic relevance. Based on the historical email titles provided by the prompt words, predict the email receiving demand; Based on the email receiving requirements, detect the percentage of spam content in the currently received emails; Based on the academic relevance and the proportion of spam content, the email receiving suggestion is output.

[0013] In the above scheme, the large model first detects the academic attributes of the email, then uses historical email titles to predict user needs, and comprehensively evaluates whether the email contains spam based on the detection results and prediction results. This makes the output result not only consider the objective content of the email, but also integrate the judgment of the user's subjective intention and the quality of the email, providing a scientific and reasonable basis for email processing strategies.

[0014] In one possible implementation of the first aspect, based on the prompt word, it is determined whether the currently received email is related to the journal, thereby obtaining the degree of academic relevance, specifically as follows: If the currently received email is related to a journal, the journal information related to the currently received email will be retrieved from existing public databases using the prompt words; wherein, the journal information includes journal field, impact factor and quartile information.

[0015] The above method uses prompts to search for publicly available journal information in order to accurately assess the academic value of emails.

[0016] In one possible implementation of the first aspect, based on the academic relevance and email reception recommendations, the currently received emails are delivered to the corresponding mailbox directory using preset email filtering criteria, specifically: The academic relevance is detected using the email filtering criteria. If the journal field and journal partition corresponding to the currently received email meet the email filtering criteria, the currently received email is delivered to the inbox. If the email receiving suggestion is set to "receive", the currently received email will be delivered to the inbox.

[0017] The above solution automatically identifies and intelligently processes currently received emails through objectively set email filtering criteria. Furthermore, it introduces email receiving suggestions, allowing emails of lower academic value to also be placed in the inbox, preventing important information from being missed due to limitations.

[0018] The second aspect of this application provides a personalized email processing system for users of the education network based on a large model. The system includes: a database construction module, a prompt word extraction module, an email information analysis module, and an email processing module. Among them, the database construction module is used to build a user's social relationship database based on paper information, communication relationships and user history operations in the previous period; The prompt word extraction module is used to extract several prompt words from the social relationship database based on the currently received email; wherein, the prompt words contain the sender's and user's historical academic information; The email information analysis module is used to predict the user's email receiving needs based on the prompt words using a preset large model, analyze the currently received emails according to the email receiving needs, and output the academic relevance of the currently received emails and email receiving suggestions. The email processing module is used to deliver currently received emails to the corresponding mailbox directory based on the academic relevance and email receiving suggestions, using preset email filtering conditions.

[0019] A third aspect of this application provides a terminal device, the device comprising: a terminal device including a processor and a memory, the memory storing a computer program, wherein the processor executes the computer program to implement the steps of the personalized email processing method for educational network users based on a large model, as described in any one of the embodiments of this application.

[0020] A fourth aspect of this application provides a storage medium storing computer-readable program code that, when executed, implements the steps of a method for processing personalized emails for users of an education network based on a large model, as described in any one of the embodiments of this application. Attached Figure Description

[0021] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram illustrating the specific process of a personalized email processing method for users of an education network based on a large model, provided in one embodiment of this application. Figure 2 This is a structural diagram of a personalized email processing system for users of an education network based on a large model, provided in one embodiment of this application. Figure 3 This is a structural diagram of a terminal device provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0025] First Embodiment In the field of anti-spam, social networks are an important source of characteristic information for determining whether an email is spam, because most business users communicate regularly with specific individuals or companies, and these communication relationships allow them to determine whether received emails are spam. However, users on educational networks differ significantly from typical business users. They frequently receive emails from strangers. These emails may be legitimate for academic exchange, such as emails discussing academic viewpoints, soliciting papers, or applying for graduate programs. These emails exhibit some common spam characteristics; for example, many soliciting papers and graduate program application emails are sent in bulk. These behaviors make such bulk emails more likely to be misclassified as spam.

[0026] Therefore, the main research objective of this application is to accurately judge received emails and filter out genuine spam in the context of educational network users, taking into account user needs and preferences.

[0027] like Figure 1 As shown, to address the problem of high false positive rates for spam in educational network user scenarios in existing technologies, the first embodiment of this application provides a detailed flowchart of a personalized email processing method for educational network users based on a large model. This embodiment's personalized email processing method for educational network users based on a large model includes steps S1 to S4, detailed below: Step S1: Construct a social relationship database for users based on paper information, communication relationships, and user history operations from the previous period.

[0028] To better filter spam and improve email detection accuracy, this application extracts multi-dimensional information, including paper information, communication relationships, and user history operations from the previous period, and constructs a social relationship database composed of multiple databases to store academic information related to emails and recent user preferences.

[0029] The social relationship database comprises five databases: a first database, a second database, a third database, a fourth database, and a fifth database. The first database stores paper information corresponding to email addresses; by retrieving the email address, one can find which papers that email address has published. The second database records all user reports and feedback, reflecting which types of emails users dislike receiving. The third database records user behavior of not reading and directly deleting emails, reflecting which types of emails users dislike receiving. The fourth database records user behavior of moving emails from the spam folder to the inbox, reflecting which types of emails users prefer to receive. The fifth database records communication records between users and various senders, reflecting which senders the user has communicated with.

[0030] The system retrieves paper information corresponding to existing email addresses from existing public databases and constructs a first database based on the mapping relationship between existing email addresses and the paper information. This first database incorporates the sender's academic identity and academic field, providing support for determining the academic value of the emails. Using this first database, the system retrieves paper information related to the email address currently receiving the email, thus obtaining the sender's published papers.

[0031] Retrieve recent user activity history from user emails, and compile statistics on user email report feedback, email deletion information, and email movement information. Construct a second, third, and fourth database based on the email addresses corresponding to these behaviors.

[0032] Among them, email reporting feedback information refers to emails that have been reported in the user's inbox, email deletion information refers to emails that have been moved directly to the spam folder without being read by the user, and email movement information refers to emails that were originally in the spam folder being moved to the inbox.

[0033] The second, third, and fourth databases can directly and accurately reflect users' preferences for email categories, revealing their preferred email types and providing a basis for subsequent personalized email processing.

[0034] The system retrieves the email addresses of historical senders from the user's email inbox, records the communication records between historical senders and the user, constructs the communication relationship between the user and historical senders, and obtains the fifth database.

[0035] Specifically, a database is built based on the email addresses of historical senders and users, recording the communication relationships between them. The database key is (sender, user), with no value. If a record can be found in the database based on the key, it means that the user has received an email from that sender.

[0036] Finally, the first, second, third, fourth, and fifth databases are integrated to construct a social relationship database for users.

[0037] Step S2: Based on the currently received email, extract several prompt words from the social relationship database.

[0038] When a user's inbox receives a new email, that email is defined as the currently received email. Extract the email address, sender information, email subject, and email body of the currently received email.

[0039] Based on the information extracted from the currently received emails, the social relationship database is searched to obtain multiple prompt words.

[0040] Based on sender and user information, extract the titles of previously published papers by the sender and user from the social relationship database. If the sender or user has not published any papers, return a null value. Extract historical sender information, as well as user reports from the previous period, email titles deleted without being read, and email titles moved from the spam folder to the inbox from the social relationship database.

[0041] The information obtained from the reports also included the email titles of the emails being reported; it also included the senders of emails that were deleted without being read and the senders of emails that were moved from the spam folder to the inbox.

[0042] The obtained email headers (including those of currently received emails), the paper title, the email body, historical sender information, and all the paper titles are combined together to form the prompt word.

[0043] The historical sender information is retrieved from the fifth database. If the social relationship database does not contain any communication relationships related to the sender, the historical sender information returned will be empty.

[0044] The prompt words clarify the current user profile information of the education network user, determine the user's preferred email type and the sender's academic information. These two pieces of information are key inputs for predicting whether the currently received emails meet the user's needs and can accurately guide the analysis direction of the large model.

[0045] Step S3: Using a preset large model, predict the user's email receiving needs based on the prompt words, analyze the currently received emails according to the email receiving needs, and output the academic relevance of the currently received emails and email receiving suggestions.

[0046] The prompt words and the currently received email are input into a pre-set large-scale model. This model has powerful logical reasoning capabilities and can determine whether the user wants to receive the currently received email based on the prompt words, thus providing corresponding email delivery suggestions. Simultaneously, the large-scale model can also evaluate the academic relevance of the currently received email based on pre-learned academic knowledge, determining whether the email is useful to the user.

[0047] The large-scale model, based on the prompt words, detects whether the currently received email is related to a journal, obtaining some academic information. If the currently received email is related to a journal, it retrieves relevant journal information from existing public databases using the prompt words, including the journal's field, impact factor, journal ranking, and the domain name of the relevant research institution. The impact factor is used to measure the journal's importance and ranking; the journal ranking information refers to the journal's ranking within a designated document and information center's journal ranking table.

[0048] Based on the richness of the academic information provided, the academic relevance of the currently received email can be predicted. If much of the academic information provided is empty (e.g., no impact factor, no journal ranking), the corresponding academic relevance is low.

[0049] The large model can also analyze users' email type preferences based on the historical email titles provided by the prompt words, in order to predict users' email receiving needs.

[0050] Based on the email receiving requirements, the body of the currently received emails is analyzed to detect whether it contains common spam content, and the percentage of spam content in the currently received emails is output.

[0051] Finally, based on the academic relevance and the proportion of spam content, an email receiving suggestion is output for the currently received emails.

[0052] Therefore, this application embodiment uses the reasoning ability of a large model to determine the academic situation of the currently received email (whether it is related to a journal), and then predicts the current user needs based on recent historical data. It comprehensively evaluates the email content, so that the output email receiving suggestion not only considers the objective content of the email, but also integrates the judgment of the user's subjective intention and the quality of the email, realizing a comprehensive evaluation of the value of the email and improving the scientificity and rationality of the decision.

[0053] Step S4: Based on the academic relevance and email receiving suggestions, the currently received emails are delivered to the corresponding mailbox directory using preset email filtering conditions.

[0054] If the email receiving suggestion is "suggest receiving", the currently received email will be delivered to the inbox; otherwise, it will be delivered to the spam folder. In addition, to improve email processing accuracy, pre-defined email filtering criteria are used to detect the academic relevance output by the large model, determining whether the currently received email is valuable to the user. If the journal field and journal category of the currently received email meet the email filtering criteria—for example, if the sender has not previously communicated with the user but has published a paper related to the user's field in a high-value journal—then the currently received email is considered to be of the type the user prefers, and the email is delivered to the inbox.

[0055] This dual-channel decision-making mechanism balances the certainty of rule enforcement with the flexibility of large-scale model judgment, enabling it to handle various complex situations more comprehensively and ensuring that users do not miss any truly important emails and receive the academic information they want in a timely manner.

[0056] Implementing the embodiments of this application has the following beneficial effects: To accurately identify academic emails, this application constructs a social relationship database based on the paper information provided in the email, the user's communication relationships, and recent user actions. This database accurately represents the user's recent preferences, social relationships, and relevant academic information, allowing the system to determine the user's specific preferences for certain emails. Hint words related to the currently received emails are extracted from the social relationship database to assist the large-scale model in determining whether the email content meets the user's needs. Furthermore, the semantic understanding capabilities of the large-scale model are used to detect whether the currently received emails contain spam, enabling more accurate identification of academic emails with potential value to the user. This provides the academic relevance of the currently received emails and email delivery suggestions, thereby automatically delivering emails to appropriate directories. This significantly improves the email processing efficiency for users on the education network, reduces the time cost of email filtering, and ensures that important academic information is not overlooked.

[0057] Second Embodiment Furthermore, in order to implement the effective reserve capacity optimization system for uncertain renewable energy output corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects, Figure 2 A structural diagram of an effective reserve capacity optimization system for addressing the uncertainty of new energy output is provided. For ease of explanation, only the parts relevant to this embodiment are shown. The effective reserve capacity optimization system for addressing the uncertainty of new energy output provided in this application embodiment includes: Database building module 201 is used to build a user's social relationship database based on paper information, communication relationships, and user history operations in the previous period.

[0058] In this embodiment of the application, paper information corresponding to existing email addresses is obtained from existing public databases, and a first database is constructed based on the mapping relationship between existing email addresses and the paper information; Based on the user's historical operations, a second database is constructed according to the user's email report feedback information, a third database is constructed according to the user's email deletion information, and a fourth database is constructed according to the user's email movement information; wherein, the email movement information refers to moving emails from the spam folder to the inbox. Based on the email addresses of historical senders and users, a communication relationship between users and historical senders is constructed, and a fifth database is constructed based on the communication relationship. The first, second, third, fourth, and fifth databases are integrated to construct the social relationship database.

[0059] The prompt word extraction module 202 is used to extract several prompt words from the social relationship database based on the currently received email; wherein, the prompt words contain the sender's and user's historical academic information.

[0060] In this embodiment, when a user's mailbox receives a new email, that email is defined as the currently received email. The email address, sender information, email title, and email body of the currently received email are extracted.

[0061] Based on the information extracted from the currently received emails, the social relationship database is searched to obtain multiple prompt words.

[0062] Based on sender and user information, extract the titles of previously published papers by the sender and user from the social relationship database. If the sender or user has not published any papers, return a null value. Extract historical sender information, as well as user reports from the previous period, email titles deleted without being read, and email titles moved from the spam folder to the inbox from the social relationship database.

[0063] The information obtained from the reports also included the email titles of the emails being reported; it also included the senders of emails that were deleted without being read and the senders of emails that were moved from the spam folder to the inbox.

[0064] The obtained email headers (including those of currently received emails), the paper title, the email body, historical sender information, and all the paper titles are combined together to form the prompt word.

[0065] The historical sender information is retrieved from the fifth database. If the social relationship database does not contain any communication relationships related to the sender, the historical sender information returned will be empty.

[0066] The prompt words clarify the current user profile information of the education network user, determine the user's preferred email type and the sender's academic information. These two pieces of information are key inputs for predicting whether the currently received emails meet the user's needs and can accurately guide the analysis direction of the large model.

[0067] The email information analysis module 203 is used to predict the user's email receiving needs based on the prompt words using a preset large model, analyze the currently received emails according to the email receiving needs, and output the academic relevance of the currently received emails and email receiving suggestions.

[0068] In this embodiment, the prompt word and the currently received email are input into a preset large-scale model. This large-scale model possesses powerful logical reasoning capabilities and can determine whether the user wants to receive the currently received email based on the prompt word, thereby obtaining corresponding email receiving suggestions. Simultaneously, the large-scale model can also evaluate the academic relevance of the currently received email based on pre-learned academic knowledge, determining whether the email is useful to the user.

[0069] The large-scale model, based on the prompt words, detects whether the currently received email is related to a journal, obtaining some academic information. If the currently received email is related to a journal, it retrieves relevant journal information from existing public databases using the prompt words, including the journal's field, impact factor, journal ranking, and the domain name of the relevant research institution. The impact factor is used to measure the journal's importance and ranking; the journal ranking information refers to the journal's ranking within a designated document and information center's journal ranking table.

[0070] Based on the richness of the academic information provided, the academic relevance of the currently received email can be predicted. If much of the academic information provided is empty (e.g., no impact factor, no journal ranking), the corresponding academic relevance is low.

[0071] The large model can also analyze users' email type preferences based on the historical email titles provided by the prompt words, in order to predict users' email receiving needs.

[0072] Based on the email receiving requirements, the body of the currently received emails is analyzed to detect whether it contains common spam content, and the percentage of spam content in the currently received emails is output.

[0073] Finally, based on the academic relevance and the proportion of spam content, an email receiving suggestion is output for the currently received emails.

[0074] Therefore, this application embodiment uses the reasoning ability of a large model to determine the academic situation of the currently received email (whether it is related to a journal), and then predicts the current user needs based on recent historical data. It comprehensively evaluates the email content, so that the output email receiving suggestion not only considers the objective content of the email, but also integrates the judgment of the user's subjective intention and the quality of the email, realizing a comprehensive evaluation of the value of the email and improving the scientificity and rationality of the decision.

[0075] The email processing module 204 is used to deliver currently received emails to the corresponding mailbox directory based on the academic relevance and email receiving suggestions, using preset email filtering conditions.

[0076] In this embodiment of the application, if the email receiving suggestion is "suggest receiving", the currently received email is delivered to the inbox; otherwise, it is delivered to the spam folder. In addition, to improve email processing accuracy, pre-defined email filtering criteria are used to detect the academic relevance output by the large model, determining whether the currently received email is valuable to the user. If the journal field and journal category of the currently received email meet the email filtering criteria—for example, if the sender has not previously communicated with the user but has published a paper related to the user's field in a high-value journal—then the currently received email is considered to be of the type the user prefers, and the email is delivered to the inbox.

[0077] This dual-channel decision-making mechanism balances the certainty of rule enforcement with the flexibility of large-scale model judgment, enabling it to handle various complex situations more comprehensively and ensuring that users do not miss any truly important emails and receive the academic information they want in a timely manner.

[0078] In some embodiments, the database construction module 201 specifically comprises: To better filter spam and improve email detection accuracy, this application extracts multi-dimensional information, including paper information, communication relationships, and user history operations from the previous period, and constructs a social relationship database composed of multiple databases to store academic information related to emails and recent user preferences.

[0079] The social relationship database comprises five databases: a first database, a second database, a third database, a fourth database, and a fifth database. The first database stores paper information corresponding to email addresses; by retrieving the email address, one can find which papers that email address has published. The second database records all user reports and feedback, reflecting which types of emails users dislike receiving. The third database records user behavior of deleting emails without reading them, reflecting which types of emails users dislike receiving. The fourth database records user behavior of moving emails from the spam folder to the inbox, reflecting which types of emails users prefer to receive. The fifth database records communication records between users and various senders, reflecting which senders the user has communicated with.

[0080] The system retrieves paper information corresponding to existing email addresses from existing public databases and constructs a first database based on the mapping relationship between existing email addresses and the paper information. This first database incorporates the sender's academic identity and academic field, providing support for determining the academic value of the emails. Using this first database, the system retrieves paper information related to the email address currently receiving the email, thus obtaining the sender's published papers.

[0081] Retrieve recent user activity history from user emails, and compile statistics on user email report feedback, email deletion information, and email movement information. Construct a second, third, and fourth database based on the email addresses corresponding to these behaviors.

[0082] Among them, email reporting feedback information refers to emails that have been reported in the user's inbox, email deletion information refers to emails that have been moved directly to the spam folder without being read by the user, and email movement information refers to emails that were originally in the spam folder being moved to the inbox.

[0083] The second, third, and fourth databases can directly and accurately reflect users' preferences for email categories, revealing their preferred email types and providing a basis for subsequent personalized email processing.

[0084] The system retrieves the email addresses of historical senders from the user's email inbox, records the communication records between historical senders and the user, constructs the communication relationship between the user and historical senders, and obtains the fifth database.

[0085] Specifically, a database is built based on the email addresses of historical senders and users, recording the communication relationships between them. The database key is (sender, user), with no value. If a record can be found in the database based on the key, it means that the user has received an email from that sender.

[0086] Finally, the first, second, third, fourth, and fifth databases are integrated to construct a social relationship database for users.

[0087] Implementing the embodiments of this application has the following beneficial effects: To accurately identify academic emails, this application constructs a social relationship database based on the paper information provided in the email, the user's communication relationships, and recent user actions. This database accurately represents the user's recent preferences, social relationships, and relevant academic information, allowing the system to determine the user's specific preferences for certain emails. Hint words related to the currently received emails are extracted from the social relationship database to assist the large-scale model in determining whether the email content meets the user's needs. Furthermore, the semantic understanding capabilities of the large-scale model are used to detect whether the currently received emails contain spam, enabling more accurate identification of academic emails with potential value to the user. This provides the academic relevance of the currently received emails and email delivery suggestions, thereby automatically delivering emails to appropriate directories. This significantly improves the email processing efficiency for users on the education network, reduces the time cost of email filtering, and ensures that important academic information is not overlooked.

[0088] Furthermore, Figure 3 This is a structural diagram of a terminal device provided in one embodiment of this application. Figure 3As shown, the terminal device 3 of this embodiment includes: at least one processor 30 (in... Figure 3 The processor 30 (only one is shown in the image) and a memory 31 and a computer program 32 stored in the memory 31 and executable on the at least one processor, wherein when the processor 30 executes the computer program 32, it can implement the steps of the personalized email processing method for educational network users based on a large model, as described in any one of the embodiments of this application.

[0089] The terminal device 3 may be a computing device such as a desktop computer, a cloud server, or a laptop computer, and the computing device may include, but is not limited to, a processor 30 and a memory 31. Figure 3 This is merely an example of terminal device 3 and does not constitute a limitation on terminal device 3. It may include more or fewer components than those shown in the figure.

[0090] This application provides a storage medium that stores computer-readable program code. When the computer-readable program code is executed, it implements the steps of the above-described method for processing personalized emails for users of an education network based on a large model.

[0091] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, or improvements made by those skilled in the art within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A large model-based educational network user personalized mail processing method, characterized by, include: Based on paper information, communication relationships, and user history operations from the previous period, a user social relationship database is constructed. Based on the currently received email, several prompt words are extracted from the social relationship database; wherein, the prompt words contain the sender's and user's historical academic information; Using a pre-defined large model, the system predicts the user's email receiving needs based on the prompt words, analyzes the currently received emails according to the email receiving needs, and outputs the academic relevance of the currently received emails and email receiving suggestions. Based on the academic relevance and email receiving suggestions, the currently received emails are delivered to the corresponding mailbox directory using preset email filtering conditions.

2. The large model-based educational network user personalized mail processing method of claim 1, wherein, The process of constructing a user's social relationship database based on paper information, communication relationships, and user history operations from the previous period specifically involves: Obtain the paper information corresponding to the existing email addresses from the existing public database, and construct a first database based on the mapping relationship between the existing email addresses and the paper information; Based on the user's historical operations, a second database is constructed according to the user's email report feedback information, a third database is constructed according to the user's email deletion information, and a fourth database is constructed according to the user's email movement information; wherein, the email movement information refers to moving emails from the spam folder to the inbox. Based on the email addresses of historical senders and users, a communication relationship between users and historical senders is constructed, and a fifth database is constructed based on the communication relationship. The first, second, third, fourth, and fifth databases are integrated to construct the social relationship database. 3.The large model-based educational network user personalized mail processing method of claim 2, wherein, Also includes: By using the first database, paper information related to the email address currently receiving the email can be obtained, thus revealing the sender's published papers. The user's preferred email type is obtained through the second, third, and fourth databases. 4.The method of claim 1, wherein, The step of extracting several prompt words from the social relationship database based on the currently received email is as follows: Extract sender information, email title, and email body from currently received emails; Based on sender and user information, extract the titles of papers published by the sender and user from the social relationship database; Extract historical sender information, email titles reported by users in the previous period, email titles of emails deleted without being read, and email titles of emails moved from the spam folder to the inbox from the social relationship database; The title of the paper, the email title, the email body, the sender history, and all the titles of the paper are used as the prompt words. 5.The large model-based educational network user personalized mail processing method of claim 1, wherein, The process involves predicting the user's email receiving needs based on the prompt words, analyzing the currently received emails according to these needs, and outputting the academic relevance of the currently received emails and email receiving suggestions. Specifically: Based on the prompt words, detect whether the currently received email is related to the journal, and obtain the academic relevance. Based on the historical email titles provided by the prompt words, predict the email receiving demand; Based on the email receiving requirements, detect the percentage of spam content in the currently received emails; Based on the academic relevance and the proportion of spam content, the email receiving suggestion is output. 6.The large model-based educational network user personalized mail processing method according to claim 5, characterized in that, The step of detecting whether the currently received email is related to the journal based on the prompt words, and obtaining the academic relevance, specifically involves: If the currently received email is related to a journal, the journal information related to the currently received email will be retrieved from existing public databases using the prompt words; wherein, the journal information includes journal field, impact factor and quartile information. 7.The large model-based educational network user personalized mail processing method of claim 1, wherein, Based on the academic relevance and email reception recommendations, the currently received emails are delivered to the corresponding mailbox directory using preset email filtering criteria, specifically: The academic relevance is detected using the email filtering criteria. If the journal field and journal partition corresponding to the currently received email meet the email filtering criteria, the currently received email is delivered to the inbox. If the email receiving suggestion is set to "receive", the currently received email will be delivered to the inbox.

8. A large model-based educational network user personalized mail processing system, characterized by, include: Database construction module, prompt word extraction module, email information analysis module, and email processing module; Among them, the database construction module is used to build a user's social relationship database based on paper information, communication relationships and user history operations in the previous period; The prompt word extraction module is used to extract several prompt words from the social relationship database based on the currently received email; wherein, the prompt words contain the sender's and user's historical academic information; The email information analysis module is used to predict the user's email receiving needs based on the prompt words using a preset large model, analyze the currently received emails according to the email receiving needs, and output the academic relevance of the currently received emails and email receiving suggestions. The email processing module is used to deliver currently received emails to the corresponding mailbox directory based on the academic relevance and email receiving suggestions, using preset email filtering conditions.

9. A terminal device, comprising: It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the steps of the personalized email processing method for educational network users based on a large model, as described in any one of claims 1 to 7.

10. A storage medium, characterized by The storage medium stores computer-readable program code, which, when executed, implements the steps of the personalized email processing method for educational network users based on a large model, as described in any one of claims 1 to 7.

Images