Cross-border e-commerce mail processing method and system based on artificial intelligence

Through the large language model based on artificial intelligence, the identification and translation of key email information, combined with historical interaction records and business rules, personalized email replies are generated, and the problems of inefficient and insufficient personalization in cross-border e-commerce email processing are solved, achieving high efficiency , accurate cross-language communication.

CN118917817AInactive Publication Date: 2024-11-08SHENZHEN MAPGOO TECH

Patent Information

Application Number
CN202411411871.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-11
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing cross-border e-commerce email processing methods have problems such as low efficiency, insufficient personalization and language barriers. Especially when facing customers of different languages ​​and cultural backgrounds, communication accuracy is difficult to guarantee and high dependence on business personnel.

Method used

A large language model based on artificial intelligence is used to identify key email information, translate and reply generation, and adjust and inspect email responses based on customer historical interaction records and business rules to ensure the personalization and professionalism of the reply.

Benefits of technology

It significantly improves communication efficiency, eliminates language barriers, enhances the personalization and professionalism of email responses, reduces manual errors and dependence, and improves customer satisfaction and trust.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917817A_ABST
    Figure CN118917817A_ABST
Patent Text Reader

Abstract

The invention provides a cross-border e-commerce mail processing method and system based on artificial intelligence. The method comprises the following steps: acquiring an original mail of a client; identifying mail key information in the original mail by using a large language model; translating the mail key information by using a large language model to obtain a translation result; obtaining a historical interaction record of the customer; the historical interaction record comprises historical demand details and historical mails; generating a mail reply draft according to the mail key information, the translation result and the historical interaction record; adjusting the mail reply draft to obtain a reply mail; checking the reply mail according to the checking rule; and replying the reply mail passing the check to the client. According to the method, the mail reply efficiency, personalization and specialty are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of automated office, and specifically relates to a cross-border e-commerce mail processing method and system based on artificial intelligence. Background Art

[0002] The rapid development of cross-border e-commerce has greatly lowered the threshold for global trade. However, as the market expands, the challenges faced by enterprises in cross-border transactions have become increasingly apparent. Enterprises need to communicate with customers in different countries and regions, which often involves different languages, time zones and cultural differences. As the main business communication tool, email plays a vital role. However, the existing email processing methods mainly include the following: 1. Manual email processing. Companies usually rely on business personnel to manually process customer emails, including reading, understanding, translating and replying. This method requires high language ability and business skills of business personnel, and the processing efficiency is limited by the working time and energy of business personnel. Manual processing is time-consuming and labor-intensive, prone to misunderstandings and delays, especially when facing customers with different languages ​​and cultural backgrounds, the accuracy of communication is difficult to guarantee. In addition, the mobility and instability of business personnel will also have a negative impact on the company's customer service quality.

[0003] 2. Email automation tools. Some companies have introduced basic email automation tools, such as email templates and automatic replies. These tools can help reduce some repetitive work and improve efficiency. However, automation tools usually have a single function and cannot handle complex customer needs and multilingual communication. In addition, the content of automatic replies usually lacks personalization, which may lead to poor customer experience and affect customer satisfaction.

[0004] 3. Third-party translation tools. In order to solve the language barrier problem, some companies use third-party translation tools or software to help business personnel translate email content from one language to another. However, existing translation tools usually cannot accurately understand the context and professional terms in business emails, and the translation results are prone to ambiguity. In addition, switching between multiple tools will also increase operational complexity and reduce overall work efficiency. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a cross-border e-commerce email processing method and system based on artificial intelligence, which improves the efficiency, personalization and professionalism of email replies.

[0006] In the first aspect, a cross-border e-commerce email processing method based on artificial intelligence includes: Get the original email from the customer; Use a large language model to identify key information in the original email; Use a large language model to translate the key information in the email to obtain the translation result; Obtain the customer's historical interaction records; historical interaction records include historical demand details and historical emails; Generate email reply drafts based on email key information, translation results, and historical interaction records; Adjust the draft email reply to get a reply email; Check the reply email according to the inspection rules; Send a reply email to the customer indicating that the inspection has passed.

[0007] Furthermore, before using the large language model to identify key information in the original email, the method further includes: Get the large language model prototype; Obtain business rules and business parameters; The large language model prototype is fine-tuned according to business rules and business parameters to obtain a large language model.

[0008] Furthermore, the key information of the email includes demand details, order details or time requirements; After using the large language model to identify key information of the original email, the method further includes: Structural storage of key email information.

[0009] Furthermore, after obtaining the translation result, the method further includes: Store the translation results and the corresponding original emails.

[0010] Furthermore, generating a draft email reply based on the key email information, translation results, and historical interaction records specifically includes: Using a large language model, we can identify the context of the original email based on the key information, translation results, and historical interaction records of the email. Generate a draft email reply covering the key information of the email based on the context information.

[0011] Furthermore, the draft email reply contains multiple to-do items; Adjustments to email reply drafts include: Receive supplementary information for each to-do item and supplement each to-do item; Configure the email style; Adjust the email reply draft based on the supplemented to-do items and email style.

[0012] Furthermore, the inspection rules include: checking whether the questions raised by the customers are responded to, checking whether the information provided by the customers is verified, whether the business measures are taken, and whether the results expected by the customers are responded to.

[0013] Furthermore, after checking the reply email according to the checking rules, it also includes: Generate inspection reports based on original emails, reply emails, and inspection rules; Use the inspection report to train and optimize the large language model.

[0014] Furthermore, after sending the reply email of passing the inspection to the customer, it also includes: Generate at least one to-do task and corresponding reminder time based on the reply email; When the reminder time of the to-do task arrives, a reminder message is generated and pushed to the business personnel.

[0015] Second, a cross-border e-commerce mail processing system based on artificial intelligence, including: Mail collection unit: used to obtain the original mails of customers; Information recognition unit: used to recognize key information in the original email using a large language model; Translation unit: used to translate key information of emails using a large language model to obtain translation results; History collection unit: used to obtain the customer's historical interaction records; historical interaction records include historical demand details and historical emails; Draft generation unit: used to generate email reply drafts based on email key information, translation results, and historical interaction records; Mail adjustment unit: used to adjust the draft of the mail reply to obtain a reply mail; Mail checking unit: used to check the reply mail according to the checking rules; Reply unit: used to send the reply email of passed inspection to the customer.

[0016] Furthermore, it also includes: Model creation unit: used to obtain a large language model prototype; obtain business rules and business parameters; fine-tune the large language model prototype according to the business rules and business parameters to obtain a large language model.

[0017] Furthermore, the key information of the email includes demand details, order details or time requirements; The information identification unit is also used to: store key information of the email in a structured manner.

[0018] Furthermore, the translation unit is also used to store the translation result and the corresponding original email.

[0019] Furthermore, the draft generation unit is specifically used to: use a large language model to identify the context information of the original email based on the key information of the email, the translation results and the historical interaction records; and generate an email reply draft covering the key information of the email based on the context information.

[0020] Furthermore, the draft reply to the email contains multiple to-do items; The email adjustment unit is also used to: receive supplementary information for each to-do item and supplement each to-do item; configure the email style; and adjust the email reply draft in combination with the supplemented to-do items and email style.

[0021] Furthermore, the inspection rules include: checking whether the questions raised by the customers are responded to, checking whether the information provided by the customers is verified, whether the business measures are taken, and whether the results expected by the customers are responded to.

[0022] Furthermore, it also includes: Report generation unit: used to generate inspection reports based on original emails, reply emails, and inspection rules; use the inspection reports to train the large language model and optimize the large language model.

[0023] Furthermore, it also includes: Reminder unit: used to generate at least one to-do task and the corresponding reminder time according to the reply email; when the reminder time of the to-do task arrives, generate reminder information and push it to the business personnel.

[0024] It can be seen from the above technical solutions that the cross-border e-commerce email processing method and system based on artificial intelligence provided by the present invention has the following advantages: 1) Significantly improve communication efficiency; by automating email reading, translation and replying, the time and energy invested in manual operations are greatly reduced, especially when processing a large number of customer emails, efficient and accurate communication can be maintained.

[0025] 2) Eliminate language barriers: Using large language models for real-time translation can ensure the accuracy of cross-language communication, reduce communication problems caused by language misunderstandings, and thus improve customer satisfaction.

[0026] 3) Enhance the personalization and professionalism of email replies; highly personalized email replies are achieved based on customers' historical interaction data and needs, improving customer trust and stickiness.

[0027] 4) Reduce human errors and dependence: Email automated inspection and optimization can reduce human errors in email replies, ensure the professionalism and appropriateness of email content, and reduce the company's over-reliance on the experience and capabilities of business personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0029] Figure 1 A flowchart of a cross-border e-commerce email processing method provided in an embodiment; Figure 2 A flowchart of a method for creating a large language model provided in an embodiment; Figure 3 A flowchart of a method for generating a draft email reply provided in an embodiment; Figure 4 A flowchart of a method for adjusting a draft email reply provided in an embodiment; Figure 5 A flowchart of a large language model optimization method provided in an embodiment; Figure 6 A flowchart of a to-do task reminder method provided in an embodiment; Figure 7 A module block diagram of a cross-border e-commerce mail processing system provided in an embodiment.

[0030] Figure 7 Among them, 1-email collection unit; 2-information recognition unit; 3-translation unit; 4-history collection unit; 5-draft generation unit; 6-email adjustment unit; 7-email checking unit; 8-reply unit; 9-model creation unit; 10-report generation unit; 11-reminder unit. DETAILED DESCRIPTION

[0031] The embodiments of the technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and the protection scope of the present invention cannot be limited thereto. It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should be the common meanings understood by those skilled in the art to which the present invention belongs.

[0032] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0033] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0034] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0035] Example: A cross-border e-commerce email processing method based on artificial intelligence, see Figure 1 ,include: S1: Get the original email from the customer; S2: Use a large language model to identify key information in the original email; S3: Use the large language model to translate the key information of the email to obtain the translation result; S4: Obtain the customer's historical interaction records; the historical interaction records include historical demand details and historical emails; S5: Generate a draft email reply based on the email key information, translation results, and historical interaction records; S6: adjusting the draft email reply to obtain a reply email; S7: Check the reply email according to the checking rules; S8: Send the reply email of passed inspection to the customer.

[0036] In this embodiment, the method receives all original emails of the customer, wherein the original emails may come from different countries or regions, and may be emails based on different languages, time zones and cultural differences. After receiving the original emails from different customers, the method calls the large language model to read and analyze the email content, and identifies the key email information in the original email. For example, after reading the email content, the large language model is called to perform natural language processing such as segmentation, keyword extraction, subject-verb-object analysis on the original email to identify the key email information in the original email. The key email information includes demand details, order details, time requirements, etc. For example, demand details include the products that the customer wants to buy and the price, etc., order details include the order number that the customer wants to check the logistics situation, etc., and the time requirements include the time when the customer wants to ship, etc. In this way, the method can accurately parse the original emails from different regions, countries, and languages, and identify the key email information of the original email.

[0037] In this embodiment, after identifying the key information of the email, the method uses a large language model to translate the key information of the email. For example, the large language model translates the key information of the email in combination with the context and industry-specific terminology. In specific implementation, the method can manually modify the language before translation (English, French, Chinese, Japanese, etc.) and the language after translation (English, French, Chinese, Japanese, etc.) to achieve accurate and real-time translation between different languages, ensure the accuracy of cross-language communication, and the translation results are more accurate compared to existing translation tools with single functions.

[0038] In this embodiment, when generating a draft email reply, the method obtains the historical interaction record of the customer. For example, the method stores each customer's original email, the historical demand details obtained after parsing the original email, and the last generated reply email in the database. When the original email of customer A is obtained, the historical demand details and historical emails corresponding to customer A are queried from the database. The historical emails include the original emails and the generated reply emails. The method generates a draft email reply based on the key information of the email, the translation results, and the historical interaction records, so that the structure of the current draft email reply is consistent with that of the historical reply email, ensuring that the reply content can continue the habit of historical interaction and ensure the personalization and professionalism of the email reply.

[0039] In this embodiment, after generating the email reply draft, the method adjusts the email reply draft to obtain a reply email. For example, the method can manually adjust each email reply draft. When adjusting the email reply draft, it can be adjusted according to the customer's demand details and business goals. The automatic + manual method can improve the accuracy of the email reply.

[0040] In this embodiment, after the method adjusts the draft email reply, the generated reply email can also be stored to facilitate subsequent process calls. After the method adjusts the draft email reply, the reply email is also checked according to the check rules. For example, the check can include grammar, semantics, politeness, professionalism, etc. to ensure the quality of the reply email. Finally, the method replies the reply email that passes the check to the customer.

[0041] This cross-border e-commerce email processing method based on artificial intelligence has the following advantages: 1) Significantly improve communication efficiency; by automating email reading, translation and replying, the time and energy invested in manual operations are greatly reduced, especially when processing a large number of customer emails, efficient and accurate communication can be maintained.

[0042] 2) Eliminate language barriers: Using large language models for real-time translation can ensure the accuracy of cross-language communication, reduce communication problems caused by language misunderstandings, and thus improve customer satisfaction.

[0043] 3) Enhance the personalization and professionalism of email replies; highly personalized email replies are achieved based on customers' historical interaction data and needs, improving customer trust and stickiness.

[0044] 4) Reduce human errors and dependence: Email automated inspection and optimization can reduce human errors in email replies, ensure the professionalism and appropriateness of email content, and reduce the company's over-reliance on the experience and capabilities of business personnel.

[0045] Further, in some embodiments, see Figure 2 Before using the large language model to identify the key information in the original email, it also includes: S11: Get the large language model prototype; S12: Obtain business rules and business parameters; S13: Fine-tune the large language model prototype according to business rules and business parameters to obtain a large language model.

[0046] In this embodiment, the large language model prototype can select the prototype of the existing large language model, for example, the large language model prototype can select the large model Llama3:70B. The method can load the large language model prototype from the cloud or computer device, and initialize the large language model prototype after the loading is successful. After initializing the large language model prototype, the method also needs to obtain business rules and business parameters, wherein the business rules and business parameters can be set in advance, or can be set during the execution of the method or input from the outside. The business rules and business parameters of different businesses are different. For example, the business rules include the email reply logic of different businesses, and the business parameters include which parameters need to be covered in the emails replied by different businesses. For example, for the order confirmation business, the business rules include that when the customer places an order, the order confirmation email needs to be sent in time; the business parameters include the order number, the number of products, the total price of the product, etc. For example, for the delivery business, the business rules include sending an email to the customer immediately after the goods are shipped to provide logistics tracking information; the business parameters include the order number, the logistics company, the logistics order number, etc. Finally, the method fine-tunes the large language model prototype according to the business rules and business parameters, so that the obtained large language model can adapt to specific business needs.

[0047] Furthermore, in some embodiments, after using the large language model to identify key email information in the original email, the method further includes: Structural storage of key email information.

[0048] In this embodiment, after the method identifies the key information of the original email, the key information of the email is stored in a structured manner, so that the key information of the email is stored in a hierarchical and organized manner, providing data support for the subsequent extraction, analysis, and processing of the key information of the email, and making data operations more convenient. For example, the key information of the email can be stored in a list manner for different customers and industries, or it can be stored in a structured manner with the help of relational data.

[0049] Furthermore, in some embodiments, after obtaining the translation result, the method further includes: Store the translation results and the corresponding original emails.

[0050] In this embodiment, after obtaining the translation result, the method can store the translation result and the corresponding original email in correspondence, so that if the business personnel need to verify whether the translation result is correct, they can call out the stored translation result and the original email, and compare the translation result and the original email to realize the manual review and verification of the translation result. The method can also train the large language model again with the translation result and the original email to optimize the large language model.

[0051] Further, in some embodiments, see Figure 3 , generating a draft email reply based on the email key information, translation results and historical interaction records, including: S21: Using a large language model, identify the contextual information of the original email based on the key information of the email, the translation results, and the historical interaction records; S22: Generate an email reply draft covering key information of the email based on the context information.

[0052] In this embodiment, the method first extracts the structure of the customer's historical reply emails, and generates the structure of the current email reply draft based on the structure of the historical reply emails, ensuring that the structure of the current reply content continues the habit of historical interaction. Then, the method can input the key information of the email and the translation results into the large language model, and use the natural language understanding ability of the large language model to accurately identify the demand details, customer expectations, context information, etc. in the original email. For example, the context information of the original email identified by the large language model is consulting production progress, inquiry, consulting processing flow, etc. One or more context information can be identified with the same original email. Finally, based on the context information, an email reply draft covering the key information of the email is generated, for example, a corresponding reply structure is generated according to each context information, and all reply structures are combined to obtain the structure of the email reply draft, and reply information is generated according to the identified demand details, customer expectations, etc., and the reply information is filled into the structure of the email reply draft to obtain an email reply draft that covers the key points of customer needs and meets expectations. The reply information can be obtained by calling a reply template or manually entering by the user. The response template can be pre-set by the management user, and the user can create a corresponding response template for each requirement detail or customer expectation.

[0053] Further, in some embodiments, the draft email reply includes multiple to-do items; See also Figure 4 , the adjustments to the draft email reply include: S31: receiving supplementary information for each to-do item, and supplementing each to-do item; S32: Configure email style; S33: Adjust the draft email reply based on the supplemented to-do items and email style.

[0054] In this embodiment, the email reply draft generated by the method may include multiple to-do items, and all to-do items constitute a to-do list. For example, the generated to-do items include quotation, order progress, etc. The to-do list covers all customer needs to ensure that each customer need can be identified and responded to. When adjusting the email reply draft, the method can supplement the existing to-do items, such as adding information or adding new to-do items to each to-do item, so that the email reply content covers the customer's needs more comprehensively. The method can also allow users to configure the email style to meet the communication needs of different scenarios. For example, the email style is concise and efficient, detailed and professional, friendly and kind, etc. The method can also allow users to configure the email tone, such as the email tone is formal and polite, friendly and kind, professional and concise, etc. Finally, the email reply draft is adjusted based on the supplemented to-do items and email style. In specific implementation, in addition to manually adjusting the email reply draft based on the supplemented to-do items and email style, this method can also input the to-do items, email style, and email tone into the large language model. The large language model adjusts the email reply draft based on the input data to ensure high consistency and professionalism in the form and content of the email.

[0055] Furthermore, in some embodiments, the checking rules include: checking whether questions raised by customers are answered, checking whether information provided by customers is verified, whether business measures are taken, and whether the results expected by customers are answered.

[0056] In this embodiment, the method can perform multi-dimensional checks on the reply email. For example, the method sets up multi-level detection rules. In addition to the dimensions of grammar, semantics, politeness, professionalism, etc., the method can also check the reply email from the following dimensions: check whether the questions raised by the customer are properly answered, check whether the information provided by the customer is verified, check whether the necessary business measures are taken, check whether the results expected by the customer are clearly answered, etc. If some of the customer's needs have not been met, during the inspection process of the reply email, appropriate explanations can also be added to the reply email to explain.

[0057] Further, in some embodiments, see Figure 5 , after checking the reply email according to the checking rules, it also includes: S41: Generate an inspection report based on the original email, the reply email, and the inspection rules; S42: Use the inspection report to train a large language model and optimize the large language model.

[0058] In this embodiment, the method can also generate a detailed inspection report based on the original email, reply email, and inspection rules for easy reference by customers. In addition, the method can also use the inspection report to continue training the large language model and optimize the large language model so that the large language model can further learn the association between the original email and the reply email, improve the reply ability and reply accuracy of the large language model, and automatically improve and optimize the large language model to ensure the accuracy, politeness, and professionalism of the email reply content.

[0059] Further, in some embodiments, see Figure 6 , after sending the reply email of passing the inspection to the customer, it also includes: S51: generating at least one to-do task and a corresponding reminder time according to the reply email; S52: When the reminder time of the to-do task arrives, a reminder message is generated and pushed to the business personnel.

[0060] In this embodiment, the method also has an email reply reminder function. The user can generate a to-do task for each reply email, wherein the generated to-do task is a to-do task that needs to be handled by the management user according to customer needs, for example, the to-do task includes following up with customers, processing orders, arranging shipments, etc. At the same time, the method also sets a reminder time according to each to-do task, and when the reminder time of the to-do task arrives, a reminder message is generated and pushed to the business personnel. For example, when a reminder event of a to-do task of following up a customer arrives, a reminder message is generated and pushed to the business personnel, reminding the business personnel to follow up with the customer to ensure the timely processing of the to-do task.

[0061] A cross-border e-commerce mail processing system based on artificial intelligence, see Figure 7 ,include: Email collection unit 1: used to obtain the original emails of customers; Information identification unit 2: used to identify key information of the original email using a large language model; Translation unit 3: used to translate key information of the email using a large language model to obtain a translation result; History collection unit 4: used to obtain the customer's historical interaction records; the historical interaction records include historical demand details and historical emails; Draft generation unit 5: used to generate an email reply draft according to the email key information, translation results and historical interaction records; Mail adjustment unit 6: used for adjusting the draft mail reply to obtain a reply mail; Mail checking unit 7: used to check the reply mail according to the checking rules; Reply unit 8: used to send a reply email of passed inspection to the customer.

[0062] Further, in some embodiments, see Figure 7 , also includes: Model creation unit 9: used to obtain a large language model prototype; obtain business rules and business parameters; and fine-tune the large language model prototype according to the business rules and business parameters to obtain a large language model.

[0063] Further, in some embodiments, the key information of the email includes demand details, order details or time requirements; The information identification unit 2 is also used to: store key information of the email in a structured manner.

[0064] Furthermore, in some embodiments, the translation unit 3 is further configured to store the translation result and the corresponding original email.

[0065] Furthermore, in some embodiments, the draft generation unit 5 is specifically used to: utilize a large language model to identify context information of the original email based on key email information, translation results, and historical interaction records; and generate an email reply draft covering the key email information based on the context information.

[0066] Further, in some embodiments, the draft reply to the email includes multiple to-do items; The email adjustment unit 6 is also used to: receive supplementary information for each to-do item, supplement each to-do item; configure the email style; and adjust the email reply draft in combination with the supplemented to-do items and email style.

[0067] Furthermore, in some embodiments, the checking rules include: checking whether questions raised by customers are answered, checking whether information provided by customers is verified, whether business measures are taken, and whether the results expected by customers are answered.

[0068] Further, in some embodiments, see Figure 7 , also includes: The report generation unit 10 is used to generate an inspection report according to the original email, the reply email, and the inspection rules; and to train the large language model using the inspection report to optimize the large language model.

[0069] Further, in some embodiments, see Figure 7 , also includes: The reminder unit 11 is used to generate at least one to-do task and a corresponding reminder time according to the reply email; when the reminder time of the to-do task arrives, a reminder message is generated and pushed to the business personnel.

[0070] The system provided in the embodiment of the present invention is briefly described. For matters not mentioned in the embodiment part, reference may be made to the corresponding contents in the aforementioned embodiment.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A cross-border e-commerce mail processing method based on artificial intelligence, characterized in that: include: Get the original email from the customer; Use a large language model to identify key information in the original email; Using a large language model to translate the key information of the email to obtain a translation result; Obtain the customer's historical interaction records; the historical interaction records include historical demand details and historical emails; Generate email reply drafts based on email key information, translation results, and historical interaction records; Adjust the draft email reply to get a reply email; Check the reply email according to the inspection rules; Send a reply email to the customer indicating that the inspection has passed.

2. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: Before using the large language model to identify the key information in the original email, it also includes: Get the large language model prototype; Obtain business rules and business parameters; The large language model prototype is fine-tuned according to business rules and business parameters to obtain the large language model.

3. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: The key information in the email includes demand details, order details or time requirements; After using the large language model to identify key email information in the original email, the method further includes: The key information of the email is stored in a structured manner.

4. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: After getting the translation result, it also includes: The translation result and the corresponding original email are stored.

5. According to claim 1, the method for processing cross-border e-commerce mails based on artificial intelligence is characterized in that: Generating an email reply draft based on email key information, translation results, and historical interaction records specifically includes: Using a large language model, we can identify the context of the original email based on the key information, translation results, and historical interaction records of the email. An email reply draft covering the key information of the email is generated based on the context information.

6. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: The draft email reply contains multiple to-do items; The adjustment of the draft email reply specifically includes: Receive supplementary information for each to-do item and supplement each to-do item; Configure the email style; Adjust the email reply draft based on the supplemented to-do items and email style.

7. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: The inspection rules include: checking whether the questions raised by the customer are answered, checking whether the information provided by the customer is verified, whether the business measures are taken, and whether the results expected by the customer are answered.

8. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: After checking the reply email according to the inspection rules, it also includes: Generate inspection reports based on original emails, reply emails, and inspection rules; The inspection report is used to train the large language model and optimize the large language model.

9. The cross-border e-commerce mail processing method based on artificial intelligence according to claim 1 is characterized in that: After sending the approval email to the customer, it also includes: Generate at least one to-do task and a corresponding reminder time according to the reply email; When the reminder time of the to-do task arrives, a reminder message is generated and pushed to the business personnel.

10. A cross-border e-commerce mail processing system based on artificial intelligence, characterized in that: include: Mail collection unit: used to obtain the original mails of customers; Information recognition unit: used to recognize key information in the original email using a large language model; Translation unit: used to translate the key information of the email using a large language model to obtain a translation result; History collection unit: used to obtain the customer's historical interaction records; the historical interaction records include historical demand details and historical emails; Draft generation unit: used to generate email reply drafts based on email key information, translation results, and historical interaction records; Mail adjustment unit: used to adjust the draft of the mail reply to obtain a reply mail; Mail checking unit: used to check the reply mail according to the checking rules; Reply unit: used to send the reply email of passed inspection to the customer.

Citation Information

Patent Citations

  • System and method for automatically generating customised respose to an email

    CA3148108A1

  • E-mail reply method and device, storage medium and computer equipment

    CN110011898A

  • Mail processing method and device combining RPA and AI and electronic equipment

    CN113315692A

Cited By

  • Business mail automatic processing method and system based on robot process automation

    CN121458250A

  • Cross-border e-mail processing method and device, computer equipment and storage medium

    CN121563454A

  • Enterprise mailbox splitting and deploying method and device fusing large model, equipment and medium

    CN121619304A