Mail reading, summarizing and replying method, system and device based on large model, and medium

Abstract

The invention discloses an email reading summary reply method, system and device based on a large model and a medium, and belongs to the technical field of email processing, and the email reading summary reply method based on the large model comprises the following steps: S100, obtaining a plurality of emails; s200, extracting the text content of each mail, generating the content summary of each mail, and further obtaining the content summary of a plurality of mails; s300, generating each mail reply according to the content summary of each mail, and further obtaining a plurality of mail replies; and step S400, according to the plurality of mail replies, judging whether each mail reply satisfies a mail sending condition, if so, sending the mail reply satisfying the mail sending condition, and if not, stopping sending the mail reply not satisfying the mail sending condition. The method and the device are suitable for batch processing of multiple mail replies.

Description

Technical Field

[0001] The present invention belongs to the field of email processing technology, and specifically relates to a method, system, device, and medium for reading, summarizing, and replying emails based on a large model. Background Art

[0002] Traditional email processing relies mainly on manual operations, and users need to manually open emails, read the content, summarize key information, and reply. This method is not only time-consuming and labor-intensive, but also prone to information omissions or misunderstandings due to human factors. Although there are some email automation tools on the market, most of them can only implement simple functions such as email forwarding and preset template replies, and are unable to deeply understand and intelligently process email content. For example, some email clients provide keyword-based filtering and automatic reply functions, but these functions are relatively limited and cannot meet the email processing needs in complex scenarios. In addition, most existing email automation tools lack the ability to understand the semantics of email content. When faced with complex email content, they are unable to accurately extract key information and generate appropriate replies. Therefore, there is an urgent need to develop new email reading, summarizing and replying technologies. Summary of the Invention

[0003] This invention provides a method, system, device, and medium for reading, summarizing, and replying emails based on a large model, aiming to partially or completely resolve the existing technical problems of reading, summarizing, and replying emails. To achieve the above objectives, this invention adopts the following technical solutions:

[0004] First, the email reading, summarizing and replying method based on a large model includes:

[0005] Step S100: Obtain multiple emails;

[0006] Step S200: extracting the text content of each email, generating a content summary of each email, and then obtaining multiple email content summaries;

[0007] Step S300: generating each email reply based on the summary of each email content, thereby obtaining multiple email replies;

[0008] Step S400: judging whether each email reply satisfies the email sending condition according to the plurality of email replies, and if so, sending the email reply that satisfies the email sending condition; if not, stopping sending the email reply that does not satisfy the email sending condition.

[0009] Optionally, step S100 includes:

[0010] Step S101: Use the mail configuration node to configure the POP mailbox protocol;

[0011] Step S102: Obtain multiple emails in the mailbox.

[0012] Optionally, step S200 includes:

[0013] Step S201: extracting the content of each email using text parsing technology;

[0014] Step S202: Use the first model to perform semantic understanding and key information extraction on the email body content to generate a summary of each email content.

[0015] Optionally, step S300 includes:

[0016] Step S301: The second model builds a reply logic structure and uses a dialogue mechanism to generate a preliminary reply to each email;

[0017] Step S302: The second largest model optimizes each preliminary email reply again, generates each email reply, and then obtains multiple email replies.

[0018] Optionally, in step S200, the first large model extracts key entities and intents, where the key entities include names, time, and places, and the intents include requests and notifications.

[0019] Optionally, in step S301, the reply logic structure includes: opening: salutation and greetings, body: key points in the reply email summary; prompts: prompts provide a clear generation goal for the second largest model, control the tone, format and content of the reply, and the second largest model adjusts the output style according to the prompts; model judgment intention: whether it is necessary to confirm receipt of the email and commit to completing the task; ending: polite closing remarks and signature.

[0020] Optionally, in step S400, the email sending conditions include: content integrity: the reply contains the salutation, body and signature, appropriate tone: the reply tone matches the email scenario, key information coverage: the reply responds to all key points of the original email, and language standardization: no spelling or grammatical errors.

[0021] In a second aspect, a large-model-based email reading summary and reply system is used to implement any of the large-model-based email reading summary and reply methods described in the first aspect, including:

[0022] Email acquisition module, acquires multiple emails;

[0023] The email summary module extracts the text content of each email and generates a content summary for each email, thereby obtaining multiple email content summaries;

[0024] The email reply module generates a reply to each email based on the summary of each email content, thereby obtaining multiple email replies;

[0025] The email sending module determines whether each email reply meets the email sending condition based on multiple email replies. If yes, the email reply that meets the email sending condition is sent. If not, the email reply that does not meet the email sending condition is stopped from being sent.

[0026] In a third aspect, a computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the method for reading, summarizing and replying emails based on a large model described in any one of the first aspects is executed.

[0027] In a fourth aspect, a large-model-based email reading summary and reply device comprises a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the large-model-based email reading summary and reply method as described in any one of the first aspects.

[0028] (1) In the present invention application, it is possible to automatically read, summarize and reply to emails, reducing manual operation time, and is particularly suitable for processing scenarios with a large number of emails, significantly improving work efficiency. The reply logic generated by the large model is clear and the content is accurate. After polishing, the language expression is more fluent and professional, improving the quality of email communication and reducing dependence on manual labor. Especially when facing repetitive and regular email processing tasks, it can significantly reduce labor costs. The large model has good adaptability to diverse email content and can flexibly respond to emails of different topics and types.

[0029] (2) In the present invention application, first of all, from email acquisition to summary, reply generation and then to sending, the entire process is highly automated, reducing manual intervention and significantly improving the efficiency of email processing. Combined with text parsing technology and large language models, the system can intelligently extract email content, generate summaries and replies, and even optimize language, adapt to complex semantics and diverse needs, judge sending conditions, ensure the quality of email replies, and reduce error rates; in addition, all modules follow unified standards (such as POP protocol, language specifications) to ensure consistent processing results and unified reply style, enhance professionalism and brand image, and each module can be independently optimized or replaced (such as replacing the large language model, adjusting the sending conditions), which is convenient for expanding functions (such as multi-language support), suitable for scenarios of different scales and needs, support manual review options, balance automation and manual control, meet the needs of sensitive or complex scenarios, and be suitable for scenarios such as corporate communications, customer service, and project collaboration. At the same time, its flexibility and scalability enable it to adapt to diverse actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 This is a flow chart of the method for reading, summarizing and replying emails based on a large model applied for by the present invention;

[0032] Figure 2 This is a schematic diagram of the composition of the large model-based email reading, summarizing and replying system applied for by the present invention;

[0033] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments; based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0035] In the description of the present invention, it should be understood that the terms "upper," "lower," "front," "back," "left," "right," "top," "bottom," "inside," and "outside" and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or components referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. To further clarify the objectives, technical solutions, and advantages of the present invention, embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0036] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means one, two, or more than two, unless otherwise specifically defined.

[0037] In order to make the purpose, technical solutions and advantages of the present invention application clearer, the technical solutions in the embodiments of the present invention application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention application. Obviously, the described embodiments are only part of the embodiments of the present invention application, not all the embodiments; based on the embodiments in the present invention application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention application.

[0038] Email reading, summarizing and replying method based on large model

[0039] like Figure 1 As shown, in the first aspect, the email reading, summarizing and replying method based on the large model includes:

[0040] Step S100: Obtain multiple emails;

[0041] Specifically, step S100 includes:

[0042] Step S101: Use the mail configuration node to configure the POP mailbox protocol;

[0043] In some embodiments, the user can open software including an email client, find the "Settings" or "Options" menu in the email client node interface, then select the "Account Settings" or "Mailbox Configuration" option, click the "Add Account" button, select the "Email Account" type, select the POP protocol, and in the POP protocol selection interface, select "POP" or "POP3" as the protocol type.

[0044] In some embodiments, enter the POP server address: enter the address provided by the email service provider, such as pop.example.com, the port number is usually 110 (unencrypted connection) or 995 (using SSL / TLS encryption), enter the mailbox username or complete email address, enter the corresponding email password, choose whether to enable SSL / TLS encryption as needed, click the "Save" or "Finish" button to complete the configuration of the POP mailbox protocol.

[0045] In some embodiments, POP (Post Office Protocol) is a standard protocol for receiving emails from a mail server. It allows the client to download emails from the server and choose whether to keep a copy of the email on the server. When the POP mailbox protocol is configured, the server address, port number, username and password provided by the user will be used by the client to establish a connection with the mail server. Security settings (such as SSL / TLS) protect data through encrypted transmission channels to avoid information leakage. After the POP mailbox protocol is configured, software including the mail client can communicate with the POP server through a TCP connection, verify the user's identity (through username and password), and prepare for subsequent email retrieval.

[0046] Step S102: Obtain multiple emails in the mailbox;

[0047] In some embodiments, software including an email client sends the following POP commands to the email server through an established connection. The POP commands include: STAT instruction: query the number and total size of emails in the email mailbox, LIST instruction: obtain a list of numbers and sizes of all emails, RETR<email number> instruction: retrieve the content of the email with a specified number, the server receives the command and returns data: returns the mailbox status, helps the client understand the number and total size of emails, lists the email numbers and sizes to facilitate the client to select emails to download, downloads the content of the specified email to the local computer, transmits the email in text format, ends the session, and the server completes the deletion operation and disconnects.

[0048] In the application of the present invention, by configuring the POP protocol, software including the mail client can establish a connection with the mail server, laying the foundation for mail acquisition, involving inputting information such as the server address, port, user name and password. The working principle relies on the connection and verification mechanism of the POP protocol. After the configuration is completed, the client uses the POP command to download multiple mails from the server, including implementing sending commands, receiving responses and storing mails. The working principle is based on the command interaction between the client and the server, and is applicable to most mail clients and server environments.

[0049] In this application, first, by configuring the POP protocol and automating email retrieval, step S100 enables batch processing of emails. Users no longer need to manually download each email, significantly saving time and effort. For example, batch downloading multiple emails at once can significantly improve work efficiency. Once configured, the system automatically executes download tasks, reducing manual intervention. Furthermore, step S100 uses the POP protocol as a standard method for email retrieval, ensuring strong compatibility. The POP protocol is a widely recognized standard in the email industry, supported by nearly all email service providers. This standardization ensures that step S100 runs smoothly on different operating systems and email clients, offering excellent cross-platform compatibility. When configuring the POP protocol, users can choose to enable SSL / TLS encryption to ensure data transmission security. This encryption prevents email content from being eavesdropped or tampered with during transmission. Furthermore, through a username and password authentication mechanism, only authorized users can access the mailbox, further enhancing security. The POP protocol allows users to flexibly manage emails, such as choosing whether to retain copies of emails on the server. Deleting emails from the server saves server storage space, while retaining copies facilitates synchronization of email data across multiple devices. This flexibility enables step S100 to adapt to different usage requirements; in addition, the operation threshold of step S100 is low, and the process of configuring the POP protocol is relatively simple. Users only need to provide basic server information and login credentials to complete the settings. In addition, the commands of the POP protocol (such as STAT, LIST, RETR) are intuitive and easy to understand, making it easy for users or developers to quickly get started and use them. The POP protocol is lightweight in design, has low requirements for system resources, has good scalability, and is easy to integrate into software systems. It is very suitable for use in resource-limited environments. Its commands and responses are concise, which reduces the use of network bandwidth, and the processing load on the client and server is also small.

[0050] Step S200: extracting the text content of each email, generating a content summary of each email, and then obtaining multiple email content summaries;

[0051] Specifically, step S200 includes:

[0052] Step S201: extracting the content of each email using text parsing technology;

[0053] In some embodiments, an email processing tool or library (such as Python's email library) is used to parse the MIME structure of the email, identify the email components, and identify the content type (Content-Type) of each part, which usually includes: header information (Headers): sender, recipient, subject, time, etc.; body content (Body): text / plain or text / html; attachments (Attachments): image, pdf, etc.

[0054] In some embodiments, the body content of text / plain type is extracted first because it is already in plain text format and does not require additional HTML parsing, reducing computing overhead. If the email only contains content of text / html type, HTML parsing is required to parse the HTML part to obtain the body content, and use HTML parsing tools (such as Python's BeautifulSoup) to process the text / html content, remove all HTML tags, scripts and styles, and only retain visible text content. Apply rules or pattern matching techniques (such as regular expressions) to identify and remove signatures (usually starting with "--"), footers, advertisements or other irrelevant content in the email, clean up extra blank lines and spaces, ensure that the output text is concise, and store the extracted plain text body as a file or variable for use in subsequent step S300.

[0055] In some embodiments, the signature block is removed: common signatures start with "--", "Best regards", "Kindregards", "Sent from my iPhone", etc., and regular expressions are used to match and truncate the signature part; footer / advertising information is removed: some email service providers will automatically add footers or advertising information (such as "From Outlook", "Thismessage was scanned"); extra blank lines and spaces are cleaned up: multiple consecutive line breaks are merged into a single line break, and the blank characters at the beginning and end of each line are deleted.

[0056] In the present invention application, the main text is effectively separated from irrelevant information such as HTML tags and signatures to ensure that the extracted content only contains core email information. It is applicable to multiple email formats (plain text or HTML), compatible with different email service providers (such as Gmail, Outlook) and clients, and filtering rules (such as signature recognition mode) can be customized according to actual needs to adapt to different languages or email styles. The extraction process can be automated through scripts or tools, which is suitable for batch processing of large amounts of emails. The parsing and cleaning process has low computational overhead and fast processing speed, making it suitable for high-throughput scenarios.

[0057] Step S202: Using the first model to perform semantic understanding and key information extraction on the email body content, and generate a summary of each email content;

[0058] In some embodiments, the extracted plain text content is summarized through a first large model. The first model identifies key entities, topics, intents, and other information in the email based on its ability to understand language and generates a content summary.

[0059] Specifically, the first model is used to perform semantic understanding and key information extraction on the plain text email content extracted in step S201 to generate a content summary of each email. Based on its language understanding ability, the first model identifies key entities (such as names, time, place), subjects, and intents (requests, notifications, etc.) in the email body content, and can write prompt words to guide the model to analyze the email content and generate a summary. For example, the prompt word can be: "Please read the following email content, extract key entities, subjects, and intents, and generate a summary of no more than 100 words."

[0060] In some embodiments, the plain text and prompt words are sent to the first large model through an API or a local interface to request generation of a summary, receive the summary returned by the first large model, check whether its format and content meet the requirements (such as length and clarity), and save the generated summary as a file, database record or variable for subsequent analysis or display.

[0061] In some embodiments, input is the plain text of an email, for example: "John, we will discuss the project progress in the conference room at 10:00 AM tomorrow. Please confirm whether you can attend."

[0062] Tip: "Analyze the following email, extract the key information, and create a concise summary."

[0063] Model call: Send the text and prompt words to the first model through the API.

[0064] Output: The first model returns a summary, such as: "John has been invited to the conference room at 10:00 AM tomorrow to discuss the project progress. Please confirm your attendance."

[0065] In some embodiments, the first large model is based on the Transformer architecture and uses a self-attention mechanism to analyze the context of the input text and understand the logical relationships and semantic structure between sentences. For example, the model can identify "tomorrow morning at 10 o'clock" as the time, "conference room" as the location, and "discussing project progress" as the topic.

[0066] In some embodiments, enter the text: Dear Alice, Please review the attached project plan and provide feedback by Friday. We will discuss the final version at the team meeting next Monday.

[0067] Tip: Please analyze the following email content, extract key entities, topics, and intent, and generate a summary of no more than 100 words:

[0068] The first model processes: key entities: Alice (name), project plan (object), Friday (time), next Monday (time), team meeting (event); judgment subject: project plan review; judgment intention: request feedback; output summary: Bob requests Alice to review the project plan attachment before Friday and will discuss the final version at the team meeting next Monday.

[0069] In some embodiments, the first large model can be pre-trained and fine-tuned to have named entity recognition (NER) and intent classification capabilities, and can extract key entities and intents. Key entities include names, time, and location, and intents include requests and notifications. For example, the first large model can determine whether an email is an "invitation" or a "confirmation request." The first large model integrates the extracted information to generate a concise natural language summary. Through generative technology, the key information is reorganized into coherent sentences, while controlling the output length and style (such as conciseness and neutrality). The prompt words provide task instructions for the model and clarify the output requirements: summary length, key information. The first large model adjusts the generation strategy based on the prompt words to ensure that the results meet user needs.

[0070] In the present invention application, first of all, the first large model can deeply understand complex semantics and extract implicit information (such as intent, key entities), going beyond simple keyword extraction. It can handle diverse email content, including unstructured or text with different language styles, and quickly generate email content summaries. It is suitable for batch processing of large numbers of emails, reducing the time for manual reading and sorting. A single API call can complete complex tasks with fast processing speed. In addition, by adjusting the prompt words or model parameters, the length, tone and degree of detail of the email content summary can be controlled to meet different needs, and support the generation of email content summaries in specific formats (such as lists and paragraphs). The output results of the first large model are unified in style, avoiding subjective differences in manual summaries and ensuring stable summary quality. Large models usually support multilingual processing, suitable for international scenarios, and can easily process emails in English, Chinese or other languages. In addition, it can handle incomplete or non-standard email content (such as missing signatures, spelling errors), and still generate reasonable email content summaries.

[0071] Step S300: generating each email reply based on the summary of each email content, thereby obtaining multiple email replies;

[0072] Specifically, step S300 includes:

[0073] Step S301: The second model builds a reply logic structure and uses a dialogue mechanism to generate a preliminary reply to each email;

[0074] In some embodiments, in step S301, based on the summary of the email content, the second large model constructs a reply logic structure, uses a dialogue mechanism to generate a preliminary reply to each email, and uses a pre-trained large language model (such as Grok, ChatGPT, LLaMA or other open source models) to generate a preliminary reply to each email through an API or local deployment call model.

[0075] From step S202, a summary of each email content is obtained as the basis for generating a reply. A structured prompt is written to clearly define the reply's goal, tone, and content requirements. The prompt may include: the email summary content, the intended reply (e.g., confirmation, answering questions, providing information), the required tone (e.g., formal, friendly), and the format requirements (e.g., including a salutation and signature). The prompt and email summary are sent to the second largest model via the API, requesting the generation of a preliminary email reply. The preliminary email reply returned by the second largest model is then checked for logic, tone, and completeness. The generated preliminary email reply is saved as a file, database record, or variable for subsequent modification or retransmission.

[0076] In some embodiments, a summary of the email content is entered, for example: "Bob requests Alice to review the attached project plan by Friday and to discuss the final version at the team meeting next Monday."

[0077] Tip: "Based on the summary of the email body below, create a formal reply email acknowledging receipt of the request, committing to completing the review on time, and expressing your anticipation for the meeting. Include a salutation and signature, and maintain a professional tone."

[0078] Model call: Send prompts and summaries to the big model through the API.

[0079] Output: A preliminary email response to the model generation process, such as: Dear Bob, Thank you for your email. I have received the attached project plan and will review it by Friday. I look forward to discussing the final version at the team meeting next Monday.

[0080] In some embodiments, the second largest model can be based on the Transformer architecture and use dialogue generation capabilities to generate natural language replies based on the input prompts and email summaries. The second largest model understands the semantics of the summary content through the self-attention mechanism and infers the logic and content of the reply. The second largest model analyzes the key information in the email content summary (such as request, time, and person) to identify the intention of the reply (such as confirmation, rejection, and providing information). For example, the second largest model can identify key information: Bob (sender), Alice (recipient), review project plan (task), Friday (deadline), and next Monday (meeting time). The second largest model can determine that "Bob requests Alice to review" requires a confirmation reply, and the second largest model generates logic: confirm receipt → promise to review on time → look forward to meeting → signature. The second-largest model can construct a reply logic structure based on the instructions in the prompt. The reply logic structure includes: opening: salutation and greetings (for example, "Thank you for your email"), body: key points in the reply email summary (for example, confirmation of review time); prompt: the prompt provides a clear generation goal for the second-largest model, controlling the tone, format and content of the reply. The second-largest model adjusts the output style (such as formal or friendly) according to the prompt; the model judges the intention: whether it is necessary to confirm receipt of the email and commit to completing the task; ending: polite closing remarks and signature.

[0081] In the present invention application, the second largest model can deeply understand the semantics of the email summary, generate replies that are consistent with the context, accurately respond to key points (such as time, tasks), handle complex or ambiguous email content, and infer appropriate reply logic. Rapidly generate reply drafts, suitable for batch processing of multiple emails, significantly reducing the time for manual reply writing, a single API call can generate a complete reply, the processing speed is fast, and by adjusting the prompts, the tone (formal, friendly), length and structure of the reply can be controlled to meet the needs of different scenarios. It supports the generation of replies in specific formats (such as containing specific phrases or signatures). The style and quality of the replies generated by the second model are stable, avoiding subjective differences or inconsistent formats in manual replies. The large model usually supports multi-language generation, suitable for processing international emails, and can easily generate replies in English, Chinese or other languages. Step S301 realizes the automatic reply generation based on the email summary through the dialogue capability of the second largest model, providing an efficient and flexible preliminary solution for the overall goal of step S302 (generating multiple email replies).

[0082] Step S302: The second largest model optimizes each preliminary email reply again, generates each email reply, and then obtains multiple email replies.

[0083] In some embodiments, step S302 polishes the preliminary reply by using the large model for a second time. For example, the pre-trained large model (such as Grok, ChatGPT, LLaMA or other open source models) can be used again to call the model through API or local deployment for text optimization, thereby improving the quality and professionalism of the reply email, providing key support for the overall goal of step S300 (generating multiple high-quality email replies), and is suitable for scenarios that require efficient and standardized email communication.

[0084] In some embodiments, a preliminary email reply is obtained from step S301 for each email, and clear prompts are written to guide the model to optimize the draft reply. The prompts typically include: optimization goals (such as improving professionalism, conciseness, and politeness), tone requirements (such as formality and friendliness), and format requirements (such as retaining the title and signature). The preliminary email reply and the prompts are sent to the second largest model via an API, requesting language polishing. The optimized preliminary email reply is returned by the second largest model as the email reply, and its sentence fluency, professionalism, and formatting completeness are checked. The email reply is saved as a file, database record, or variable for subsequent transmission or archiving.

[0085] In some embodiments, the initial email reply reads: Dear Bob, I received your email. I'll review the project plan and give you feedback by Friday. See you at the meeting next Monday. Best wishes, Alice.

[0086] The prompt reads: "Optimize the following initial email reply to be more professional, concise, and polite, while preserving the formal tone and original structure (salutation, signature, etc.)." The initial email reply and prompt are sent to the second-largest model via the API; the second-largest model outputs the email reply: Dear Bob, thank you for your email. I have received the project plan and will ensure I review and provide feedback by Friday. I look forward to discussing it with you in our meeting next Monday. Best wishes, Alice!

[0087] In some embodiments, the second largest model can be based on the Transformer architecture and use its natural language generation capabilities to analyze the semantics, grammar and tone of the initial email reply. The second largest model identifies parts that can be improved (such as lengthy expressions, less formal vocabulary) and generates more fluent and professional alternative texts. Sentence structure adjustment: The second model optimizes the sentence structure to make it clearer and more logically coherent. For example, merge scattered short sentences into fluent long sentences, or simplify complex sentences into more concise expressions. Vocabulary selection optimization: The second largest model replaces inappropriate or overly colloquial words and selects terms that are more in line with professional email standards. For example, replace "I will look at it" with "I will review it" and expand "See you at the meeting next Monday" to "Looking forward to discussing it with you at the meeting next Monday." Tone control: According to the prompt requirements, the second largest model adjusts the tone of the reply (such as from casual to formal) to ensure that the tone matches the relationship with the recipient and the email scenario. The prompts provide optimization goals and constraints for the second largest model, clarifying the aspects that need to be improved (such as professionalism and conciseness) and the structures that should be retained (such as titles and signatures). The second largest model generates optimized text that meets the requirements based on the prompts.

[0088] In the present invention application, first, the optimized initial email reply uses more formal vocabulary and structure, which conforms to the expression standards of professional emails, enhances the credibility and professional image of the email, and is suitable for business scenarios or occasions requiring formal communication. The optimized sentence structure is clearer and smoother, and the logic is more coherent, which facilitates the recipient to quickly understand the reply content and avoids the lengthy or scattered expressions that may appear in the initial email reply. The second largest model improves the tone by adding greetings or polite closing remarks, making the reply more friendly and appropriate, suitable for maintaining good communication relationships; in addition, it quickly completes language polishing, which is suitable for batch processing of multiple email replies and reducing the time for manual modification. A single API call can generate high-quality optimized text. By adjusting the prompts, the degree of optimization (such as more concise or more detailed), tone (such as more formal or more friendly) and format can be controlled, supporting the generation of replies that conform to specific cultural or industry standards. The second largest model ensures that all optimized replies have a unified style, avoiding differences in tone or format caused by manual modification, and is suitable for large-scale email processing scenarios.

[0089] Step S400: judging whether each email reply satisfies the email sending condition according to the plurality of email replies, and if so, sending the email reply that satisfies the email sending condition; if not, stopping sending the email reply that does not satisfy the email sending condition.

[0090] In some embodiments, all email replies are collected from step S302 and email sending conditions are set. The email sending conditions include: content completeness: the reply contains the salutation, body and signature, appropriate tone: the reply tone matches the email scenario (such as formal or friendly), key information coverage: the reply responds to all key points of the original email, and language standardization: no spelling or grammatical errors.

[0091] In some embodiments, email replies are displayed on a software system including an email client and an email review and sending terminal for manual review or review using a software system. The manual reviewer or the software system reviews and confirms whether the email sending conditions are met, and marks "passed" or "failed". For email replies that meet the email sending conditions, the email sending tool is used to automatically send them through the SMTP protocol. For email replies that do not meet the email sending conditions, the reason is recorded (such as "missing key information"), archived or marked as pending modification, and not sent.

[0092] In some embodiments, for manual review or machine review, content completeness: includes salutation, body, and signature (passed), tone: formal, suitable for work scenarios (passed), key information: responds to review requests and meeting arrangements (passed), language standardization: no spelling or grammatical errors (passed), results: meets the email sending conditions, automatically sent to Bob's email address, does not meet the email sending conditions, for example, if the email reply lacks a signature or does not respond to key requests (such as not mentioning the Friday deadline), it is marked as "failed" and stops sending.

[0093] In some embodiments, the manual review interface displays the email reply, the manual reviewer manually confirms it according to the conditions, the software system records the review results, and updates the reply status; or, the software system evaluates whether the email reply meets the email sending conditions based on preset rules (such as regular expression checking signatures, keyword matching checking key information) or semantic analysis of a large language model, uses the SMTP protocol to establish a connection with the email server, and sends email replies that meet the email sending conditions. The software system can maintain the status of each email reply (pending review, passed, failed, sent), and record the operation history through logs. If the sending fails (for example, a server connection problem), the system records the error and retries or notifies the administrator. Replies that do not meet the conditions are isolated to prevent accidental sending.

[0094] In some embodiments, a reply is entered, such as: Dear Bob, Thank you for your email. I have received the attached project plan and will review and provide feedback by Friday. I look forward to discussing it with you in our meeting next Monday. Best wishes, Alice! The above email reply includes a salutation, a body, and a signature, responding to the review request and meeting arrangement. The tone is formal, marked as approved, and sent to Bob's email address via the SMTP server.

[0095] In the present invention application, through machine review or manual review, it is ensured that only high-quality, compliant replies are sent, avoiding the accidental sending of incomplete or unprofessional replies, improving the reliability and professional image of email communication, and automatically judging and sending processes to reduce manual intervention. It is suitable for batch processing of large numbers of email replies, quickly completing condition checking and sending, saving time, and email sending conditions can be customized according to needs (such as tone requirements in different scenarios). It supports combination of manual review or automatic review to adapt to email processing tasks of different complexities, stops sending replies that do not meet the conditions, and prevents sending erroneous or inappropriate emails. Log records facilitate tracking problems and are easy to integrate into larger email processing platforms. It supports docking with other automation tools (such as OA systems). Manual review can provide additional security for sensitive emails, provides a reliable closed loop for the entire email processing process, and improves the efficiency and accuracy of automated email replies.

[0096] Email reading, summarizing and replying system based on large model

[0097] like Figure 2 As shown, in the second aspect, a large model-based email reading summary reply system is used to implement any large model-based email reading summary reply method described in the first aspect, including:

[0098] It should be noted that the large-scale model-based email reading summary and reply system applied for in the present invention is used to implement any of the large-scale model-based email reading summary and reply methods in the first aspect, and accordingly also includes: all the technical problems, technical solutions and technical effects recorded in any of the large-scale model-based email reading summary and reply methods in the first aspect, which will not be repeated here in the present invention application.

[0099] Email acquisition module, acquires multiple emails;

[0100] The email summary module extracts the text content of each email and generates a content summary for each email, thereby obtaining multiple email content summaries;

[0101] The email reply module generates a reply to each email based on the summary of each email content, thereby obtaining multiple email replies;

[0102] The email sending module determines whether each email reply meets the email sending condition based on multiple email replies. If yes, the email reply that meets the email sending condition is sent. If not, the email reply that does not meet the email sending condition is stopped from being sent.

[0103] In some embodiments, the email acquisition module acquires the specific contents of multiple emails, which may be the same as all the technical problems, technical solutions, and technical effects recorded in step S100, and the present application will not elaborate on this.

[0104] In some embodiments, the email summary module extracts the text content of each email, generates a content summary of each email, and then obtains the specific content of multiple email content summaries, which can be the same as all the technical problems, technical solutions, and technical effects recorded in step S200. The present invention application will not go into details about this.

[0105] In some embodiments, the email reply module generates each email reply based on the summary of each email content, and then obtains the specific content of multiple email replies, which can be the same as all the technical problems, technical solutions, and technical effects recorded in step S300. The present invention application will not go into details about this.

[0106] In some embodiments, based on multiple email replies, it is determined whether each email reply meets the email sending conditions. If so, the email reply that meets the email sending conditions is sent; if not, the sending of the email replies that do not meet the email sending conditions is stopped. The specific content may be the same as all the technical problems, technical solutions, and technical effects recorded in step S400, and the present invention application will not go into details about this.

[0107] In the present invention application, first of all, from email acquisition to summary, reply generation and then to sending, the entire process is highly automated, reducing manual intervention and significantly improving the efficiency of email processing. Combined with text parsing technology and large language models, the system can intelligently extract email content, generate summaries and replies, and even optimize language, adapt to complex semantics and diverse needs, judge sending conditions, ensure the quality of email replies, and reduce error rates; in addition, all modules follow unified standards (such as POP protocols, language specifications) to ensure consistent processing results and unified reply styles, enhance professionalism and brand image, and each module can be independently optimized or replaced (such as replacing large language models, adjusting sending conditions), which is convenient for expanding functions (such as multi-language support), suitable for scenarios of different scales and needs, support manual review options, balance automation and manual control, meet the needs of sensitive or complex scenarios, and are suitable for scenarios such as corporate communications, customer service, project collaboration, etc. At the same time, its flexibility and scalability enable it to adapt to diverse actual needs.

[0108] Computer-readable storage medium

[0109] In a third aspect, the present invention application provides a computer-readable storage medium having instructions stored thereon. When the instructions are executed on a computer, the large model-based email reading, summarizing and replying method as described in any one of the first aspects is executed.

[0110] Email reading, summarizing and replying device based on large model

[0111] In the fourth aspect, the present invention application provides an email reading summary and reply device based on a large model, including a memory and a processor that are communicatively connected, the memory is used to store a computer program, and the processor is used to read the computer program and execute the email reading summary and reply method based on a large model as described in any one of the first aspects.

[0112] In some embodiments, a large-model-based email reading summary and reply apparatus includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the large-model-based email reading summary and reply methods described in the first aspect are implemented.

[0113] Those skilled in the art will understand that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

[0114] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0115] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0116] Wherein, if the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention applies to implement all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0117] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0118] The above is a preferred embodiment of the present invention. It should be pointed out that ordinary technicians in this technical field can make several improvements and modifications without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for reading, summarizing and replying emails based on a large model, characterized in that: include: Step S100: Obtain multiple emails; Step S200: extracting the text content of each email, generating a content summary of each email, and then obtaining multiple email content summaries; Step S300: generating each email reply based on the summary of each email content, thereby obtaining multiple email replies; Step S400: judging whether each email reply satisfies the email sending condition according to the plurality of email replies, and if so, sending the email reply that satisfies the email sending condition; if not, stopping sending the email reply that does not satisfy the email sending condition.

2. The method for reading, summarizing and replying emails based on a large model according to claim 1, characterized in that: Step S100 includes: Step S101: Use the mail configuration node to configure the POP mailbox protocol; Step S102: Obtain multiple emails in the mailbox.

3. The method for reading, summarizing and replying emails based on a large model according to claim 2, characterized in that: Step S200 includes: Step S201: extracting the content of each email using text parsing technology; Step S202: Use the first model to perform semantic understanding and key information extraction on the email body content to generate a summary of each email content.

4. The method for reading, summarizing and replying emails based on a large model according to claim 1, characterized in that: Step S300 includes: Step S301: The second model builds a reply logic structure and uses a dialogue mechanism to generate a preliminary reply to each email; Step S302: The second largest model optimizes each preliminary email reply again, generates each email reply, and then obtains multiple email replies.

5. The method for reading, summarizing and replying emails based on a large model according to claim 1, characterized in that: In step S200, the first large model extracts key entities and intents, where the key entities include names, time, and location, and the intents include requests and notifications.

6. The method for reading, summarizing and replying emails based on a large model according to claim 4, characterized in that: In step S301, the reply logic structure includes: opening: salutation and greetings, body: key points in the reply email summary; prompt: the prompt provides a clear generation goal for the second largest model, controls the tone, format and content of the reply, and the second largest model adjusts the output style according to the prompt; the model determines the intention: whether it is necessary to confirm receipt of the email and commit to completing the task; ending: polite closing remarks and signature.

7. The method for reading, summarizing and replying emails based on a large model according to claim 6, characterized in that: In step S400, the email sending conditions include: content integrity: the reply contains the salutation, body and signature, appropriate tone: the reply tone matches the email context, key information coverage: the reply responds to all key points of the original email, and language standardization: no spelling or grammatical errors.

8. A system for summarizing and replying emails based on a large model, for implementing the method for summarizing and replying emails based on a large model according to any one of claims 1 to 7, characterized in that: include: Email acquisition module, acquires multiple emails; The email summary module extracts the text content of each email and generates a content summary for each email, thereby obtaining multiple email content summaries; The email reply module generates a reply to each email based on the summary of each email content, thereby obtaining multiple email replies; The email sending module determines whether each email reply meets the email sending condition based on multiple email replies. If yes, the email reply that meets the email sending condition is sent. If not, the email reply that does not meet the email sending condition is stopped from being sent.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions, and when the instructions are executed on the computer, the method for reading, summarizing and replying emails based on a large model according to any one of claims 1 to 7 is executed.

10. A large-model-based email reading summary and reply device, comprising a memory and a processor that are communicatively connected, wherein the memory is used to store a computer program, and the processor is used to read the computer program and execute the large-model-based email reading summary and reply method described in any one of claims 1 to 7.

Images