Instant messaging method and device of cross-border e-commerce system, storage medium and terminal

Abstract

The invention discloses an instant messaging method and device for a cross-border e-commerce system, a storage medium and a terminal, and the method comprises the steps: receiving an original text which is intended to be sent to an opposite terminal by a user, and employing a first language to express the original text; a prompt word is determined according to the information of the opposite terminal, and in the first instant messaging session with the opposite terminal, the prompt word at least comprises the IP address of the opposite terminal; the original text and the cue words are input into a preset large language model to obtain a target text, the preset large language model is used for translating the original text into the target text according to the cue words, the target text is expressed by adopting a second language, and the second language is determined at least according to the cue words; and presenting the target text on a session interface with the opposite terminal, and sending the target text to the opposite terminal in response to a confirmation instruction of the user. Through the scheme of the invention, the communication efficiency and the communication accuracy of a cross-border e-commerce scene can be improved, and the consultation experience of customers is optimized.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to an instant communication method and device, a storage medium, and a terminal of a cross-border e-commerce system. Background Art

[0002] In their journey towards cross-border expansion, domestic companies often face a thorny challenge: because the languages ​​used by consumers at different sites in different countries vary greatly, it is difficult for companies to quickly and effectively handle customer inquiries and respond in a timely manner when handling pre-sales and after-sales inquiries.

[0003] In the traditional customer service system, when overseas customers (hereinafter referred to as customers) provide pre-sales / after-sales consultation, the customer service will translate the reply into the customer's language and reply to the customer. This processing mechanism is highly dependent on the customer service's personal language level, and is prone to situations such as stiff translation and "Chinese translation", which affects the customer's consultation experience. In serious cases, it may cause reading difficulties for customers and cause ambiguity.

[0004] In addition, for customers who are communicating for the first time, customer service needs to wait until the customer asks a question before confirming the language used by the customer. This passive response mechanism will cause customer service to respond lags and affect communication efficiency. If customer service actively sends content before the customer speaks for the first time, since the customer's language cannot be determined, the language used by customer service to send content may not match the actual language used by the customer, affecting communication accuracy. Summary of the invention

[0005] The technical problem solved by the present disclosure is how to improve the communication efficiency and accuracy in cross-border e-commerce scenarios.

[0006] To solve the above technical problems, an embodiment of the present disclosure provides an instant messaging method for a cross-border e-commerce system, comprising: receiving an original text that a user intends to send to a peer end, the original text being expressed in a first language; determining a prompt word according to information of the peer end, wherein, in the first instant messaging session with the peer end, the prompt word at least includes the IP address of the peer end; inputting the original text and the prompt word into a preset large language model to obtain a target text, the preset large language model being used to translate the original text into a target text according to the prompt word, the target text being expressed in a second language, the second language being determined at least according to the prompt word; presenting the target text on a session interface with the peer end, and sending the target text to the peer end in response to a confirmation instruction of the user.

[0007] Optionally, in a non-first instant messaging session with the peer, the prompt word further includes: a user feature of the peer, the user feature is extracted from historical session content of the peer; wherein the second language is determined based on the user feature.

[0008] Optionally, the prompt word also includes: a language habit indicator, used to indicate the language habit of the other party; a conversation intention indicator, used to indicate the conversation intention of the other party; an expression style indicator, used to indicate the expression style of the user; and a user personality indicator, used to indicate the user's personality.

[0009] Optionally, in a single instant messaging session with the opposite end, the prompt word further includes: a temporary language indicator, the temporary language indicator is determined according to the previous context of the current instant messaging session; wherein the second language is determined according to the temporary language indicator.

[0010] Optionally, presenting the target text on the session interface with the peer end includes: presenting the target text in a preview area of ​​the session interface; in response to a confirmation instruction of the user, sending the target text to the peer end includes: presenting the target text in a session area of ​​the session interface in response to receiving the confirmation instruction; wherein, the target text presented in the preview area is not visible to the peer end, and the target text presented in the session area is visible to the peer end.

[0011] Optionally, after presenting the target text in the preview area of ​​the conversation interface, the method further includes: receiving a first instruction, the first instruction being used to update the prompt word; inputting the original text and the updated prompt word into the preset large language model to obtain an updated target text; and presenting the updated target text in the preview area.

[0012] Optionally, after the target text is presented in the preview area of ​​the conversation interface, the method further includes: receiving a second instruction, the second instruction being used to update a prompt word and indicate a target character, the target character being extracted from the target text; inputting at least the original character corresponding to the target character and the updated prompt word into the preset large language model to obtain an updated target text, the original character being extracted from the original text; and presenting the updated target text in the preview area.

[0013] Optionally, after presenting the target text in the preview area of ​​the conversation interface, the method further includes: in response to receiving a third instruction, presenting a verification text in the preview area, the verification text being the target text expressed in the first language.

[0014] To solve the above technical problems, the embodiments of the present disclosure also provide an instant messaging device for a cross-border e-commerce system, comprising: a receiving module, used to receive an original text that a user intends to send to a peer end, the original text being expressed in a first language; a processing module, used to determine a prompt word according to information of the peer end, wherein, in the first instant messaging session with the peer end, the prompt word at least includes the IP address of the peer end; a generating module, used to input the original text and the prompt word into a preset large language model to obtain a target text, the preset large language model being used to translate the original text into a target text according to the prompt word, the target text being expressed in a second language, the second language being determined at least according to the prompt word; a feedback module, used to present the target text on a session interface with the peer end, and in response to a confirmation instruction of the user, send the target text to the peer end.

[0015] To solve the above technical problems, an embodiment of the present disclosure further provides a computer-readable storage medium, which is a non-volatile storage medium or a non-transient storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are executed.

[0016] To solve the above technical problems, an embodiment of the present disclosure further provides a terminal, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and the processor executes the steps of the above method when running the computer program.

[0017] In order to solve the above technical problems, the embodiments of the present disclosure further provide a computer program product, including a computer program / instruction, which implements the steps of the above method when executed by a processor.

[0018] Compared with the prior art, the technical solution of the embodiment of the present disclosure has the following beneficial effects:

[0019] The embodiment of the present disclosure provides an instant messaging method for a cross-border e-commerce system, comprising: receiving an original text that a user intends to send to a peer end, the original text being expressed in a first language; determining a prompt word according to information of the peer end, wherein, in a first instant messaging session with the peer end, the prompt word at least includes an IP address of the peer end; inputting the original text and the prompt word into a preset large language model to obtain a target text, the preset large language model being used to translate the original text into a target text according to the prompt word, the target text being expressed in a second language, the second language being determined at least according to the prompt word; presenting the target text on a session interface with the peer end, and sending the target text to the peer end in response to a confirmation instruction of the user.

[0020] The prior art requires users to passively wait for the other party of the instant messaging session to send out a consultation question before they can determine the language used by the other party based on the expression of the consultation question. In comparison, this implementation automatically identifies the language used by the other party (i.e., the second language) based on the IP address of the other party. As a result, the language used by the other party can be accurately identified at the first time the instant messaging session is established, ensuring that the instant messaging session sent by the user to the other party is always presented in the language accustomed to by the other party, thereby improving communication accuracy. Furthermore, the language of the other party is preset based on the IP address, and after actually receiving the content sent by the other party (e.g., consultation questions), the user can promptly feedback the reply content to the other party in the second language, thereby improving communication efficiency.

[0021] The existing technology requires users to translate the reply content (i.e., original text) into a second language and feed it back to the other party, resulting in poor communication accuracy. In comparison, the present implementation scheme realizes automatic translation of text by presetting a large language model, avoiding situations such as stiff translation and Chinese-style translation, ensuring that the target text presented to the other party can accurately express the meaning of the original text expressed by the user in the first language and is easy for the other party to understand, thereby improving communication accuracy.

[0022] This implementation scheme can be applied to cross-border e-commerce scenarios, where the other end can be an overseas customer using the cross-border e-commerce system, and the user can be a domestic customer service using the cross-border e-commerce system. The other end and the user implement cross-border e-commerce business through the cross-border e-commerce system. By accurately and automatically identifying the second language based on the IP address, and accurately translating the original text input by the user into the target text in combination with the preset large language model, the customer service reception efficiency can be improved and the customer's consulting experience can be enhanced. For example, in a product consulting scenario, the other end can be an overseas consumer, and the user can be the customer service of the brand to which the product belongs. The use of this implementation scheme can enhance the consumer's brand pre-sales / after-sales experience.

[0023] Furthermore, in a non-first instant messaging session with the peer, the prompt word also includes: user features of the peer, the user features are extracted from the historical conversation content of the peer; wherein the second language is determined based on the user features. Thus, the idiomatic language of the peer is further accurately identified in combination with the historical conversation, ensuring that the target text converted by the preset large language model can be accurately understood by the peer.

[0024] Furthermore, the prompt words also include: a language habit indicator for indicating the language habit of the other party; a conversation intention indicator for indicating the conversation intention of the other party; an expression style indicator for indicating the expression style of the user; and a user persona indicator for indicating the user's persona. As a result, the target text output by the preset large language model can be more in line with what the user really wants to express, rather than a mechanical translation, and the final presentation is more in line with the language habits of the other party, improving communication accuracy and instant messaging experience.

[0025] Furthermore, in a single instant messaging session with the peer, the prompt word also includes: a temporary language indicator, the temporary language indicator is determined according to the previous text of the current instant messaging session; wherein the second language is determined according to the temporary language indicator. In a cross-border e-commerce scenario, the same account / device of the peer may be managed by different people, and thus the language used by the peer may change in a single instant messaging session. This embodiment can automatically identify the change and generate a temporary language indicator. The preset large language model timely adjusts the second language used in the output target text according to the temporary language indicator, so that the second language used in the target text sent to the peer on the conversation interface can automatically change with the change of the language used by the user of the peer. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of an instant messaging method for a cross-border e-commerce system according to an embodiment of the present disclosure;

[0027] Figures 2 to 6 are various exemplary schematic diagrams of a conversation interface in a typical application scenario of an embodiment of the present disclosure;

[0028] Figure 7 It is a structural schematic diagram of an instant messaging device of a cross-border e-commerce system according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] In order to make the above-mentioned objects, features and beneficial effects of the present disclosure more obvious and easy to understand, the specific embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0030] Figure 1 It is a flow chart of an instant communication method of a cross-border e-commerce system according to an embodiment of the present disclosure.

[0031] This implementation scheme can be applied to cross-border e-commerce scenarios, that is, customers (e.g., consumers) and users (e.g., customer service of the brand of the product) can conduct cross-border business through the cross-border e-commerce system. Customers and users are in different countries / regions, and accordingly, the language used by customers and the language used by users may be different. For example, customers speak English and users speak Chinese, where the language used by customers / users can be the native language of customers / users or the language that customers / users use more frequently in daily life. Next, we will take the example of overseas customers and domestic users for detailed explanation. The language used by the user is recorded as the first language, and the language used by the customer is recorded as the second language. The first language and the second language are different languages.

[0032] The cross-border e-commerce system can provide a cross-border e-commerce platform, and overseas customers and domestic users can conduct cross-border e-commerce business through the cross-border e-commerce system. For example, domestic users can publish products in the cross-border e-commerce system, and overseas customers can browse products and conduct transactions in the cross-border e-commerce system.

[0033] During browsing and transaction operations, overseas customers can communicate with domestic users through the cross-border e-commerce system for pre-sales consultation and / or after-sales consultation. For example, overseas customers can initiate an instant messaging session in the cross-border e-commerce system, and overseas customers and domestic users can send messages to each other in the session interface of the instant messaging session to communicate instantly. For another example, domestic users can also actively initiate an instant messaging session with overseas customers.

[0034] In this implementation, the local end (also called user) of the instant messaging session corresponds to a domestic user, and the other end of the instant messaging session corresponds to an overseas customer. This implementation can be executed by a terminal on the user side. The terminal on the user side can be deployed with a cross-border e-commerce system, and the terminal can be integrated or coupled with a display module, and the display interface of the display module is used to present the session interface of the instant messaging session.

[0035] Specifically, refer to Figure 1 The instant communication method of the cross-border e-commerce system described in this implementation scheme may include the following steps (abbreviated as S):

[0036] S101, receiving an original text that a user intends to send to a peer end, wherein the original text is expressed in a first language;

[0037] S102, determining a prompt word according to the information of the peer end, wherein, in the first instant messaging session with the peer end, the prompt word at least includes the IP address of the peer end;

[0038] S103, inputting the original text and the prompt word into a preset large language model to obtain a target text, wherein the preset large language model is used to translate the original text into a target text according to the prompt word, wherein the target text is expressed in a second language, and the second language is determined at least according to the prompt word;

[0039] S104, presenting the target text on the conversation interface with the opposite end, and sending the target text to the opposite end in response to a confirmation instruction from the user.

[0040] More specifically, the IP address of the other end may be the Internet Protocol address of the network device used by the other end for this instant messaging session. For example, in response to the other end or the user establishing this instant messaging session, the terminal may obtain the IP address of the other end by interacting with the network device used by the other end.

[0041] Furthermore, if this instant messaging session is the first instant messaging session between the user and the other party, that is, the user and the other party have not had any instant messaging interaction in the cross-border e-commerce system in the past, the terminal obtains the IP address of the other party after the instant messaging session is established. Furthermore, in response to obtaining the IP address of the other party, the country (or region) where the other party is currently located can be determined, and then the common language of the country (or region) can be determined as the second language.

[0042] In some embodiments, a preset large language model may be pre-trained using a training set, and the training set may include multiple IP addresses, the country (or region) to which each IP address belongs, and the common language corresponding to each country (or region). Accordingly, the trained preset large language model may determine the corresponding second language according to the input IP address. Thus, the IP address may be input as one of the prompt words into the preset large language model, so that the preset large language model may determine the second language to be used in this instant messaging session.

[0043] Furthermore, the translation capability of a preset large language model can be trained in advance using massive text data, so that the trained preset large language model can smoothly convert the input original text from a first language of any language type into a second language determined based on the prompt word.

[0044] Taking the cross-border e-commerce customer service scenario as an example, in response to the other party sending the consultation content on the conversation interface, the user can enter the reply content (i.e., the original text) in the first language in the input box of the conversation interface and click the "Translate" button of the conversation interface. In response to the "Translate" button being triggered, the terminal executes S101 to obtain the original text, and executes S102 to determine the prompt word containing at least the IP address of the other party, and then the terminal can input the original text and the prompt word into the preset large language model together. The preset large language model automatically translates the original text into the second language determined according to the prompt word and outputs it. In response to receiving the target text output by the preset large language model, the terminal can present the target text in the preview area of ​​the conversation interface for the user to browse. The user can click the "Apply" button of the conversation interface to issue a confirmation instruction. In response to receiving the confirmation instruction, the terminal sends the target text to the other party in the conversation interface.

[0045] In some embodiments, the order of executing S102 and S101 is not limited. For example, S102 can be executed at the same time as the instant messaging session is established, so as to determine the second language at the first time and improve the subsequent communication efficiency. For another example, S102 can be executed at any time during the instant messaging session, so that the prompt words can be updated in real time during the instant messaging process with the other party to improve the communication accuracy.

[0046] In some embodiments, the preset large language model can be deployed in the cross-border e-commerce system so as to be called when needed. In a variation, the preset large language model can also be a general model integrated with a third party, and the terminal can access the preset large language model when needed to execute this embodiment.

[0047] In a specific implementation, in a non-first instant messaging session with the other party, the prompt word may also include user features of the other party, the user features are extracted from historical conversation content of the other party; wherein the second language is determined based on the user features.

[0048] Specifically, if this instant messaging session is not the first instant messaging session between the user and the other party, that is, the user and the other party have had instant messaging interactions in the cross-border e-commerce system in the past, then after this instant messaging session is established, in addition to obtaining the IP address of the other party, the terminal can also obtain historical conversation content with the other party.

[0049] Furthermore, by analyzing the historical conversation content of the other party using natural language processing and other means, the second language used by the other party can be more accurately identified. Furthermore, the user characteristics may include the language type with the highest proportion of use in the historical conversation content.

[0050] Furthermore, the language type and IP address used in the historical conversation content with the highest proportion are both input as prompt words into the preset large language model, and the preset large language model can assign a first weight to the language type used in the historical conversation content with the highest proportion, and assign a second weight to the language type determined based on the IP address, wherein the first weight is higher than the second weight. In response to the fact that the language type used in the historical conversation content with the highest proportion is different from the language type determined based on the IP address, the preset large language model determines the language type used in the historical conversation content with the highest proportion, which has a higher weight, as the target language (i.e., the second language) for this translation.

[0051] For example, assuming that the language type determined based on the IP address of the other party is English, and the language type most commonly used in the historical conversation content of the other party is French, the preset large language model preferably converts the input original text expressed in Chinese into French rather than English.

[0052] Therefore, the idiomatic language of the other party can be further accurately identified in combination with historical conversations, ensuring that the target text converted by the preset large language model can be accurately understood by the other party.

[0053] In a specific implementation, the prompt word may also include a language habit indicator for indicating the language habit of the other party; a conversation intention indicator for indicating the conversation intention of the other party; and an expression style indicator for indicating the expression style of the user.

[0054] Specifically, the language habits of the other party can also be obtained based on the analysis of the historical conversation content of the other party. For example, natural language processing is performed on the historical conversation content to understand the expression habits of the other party. Language habits may include idiomatic words, catchphrases, commonly used punctuation marks, sentence segmentation styles, etc.

[0055] Inputting the language habit indicator as one of the prompt words into the preset large language model helps the preset large language model to make the generated target text more consistent with the language habits of the other party when converting the original text into the target text.

[0056] Furthermore, the intention of the other party to conduct this instant messaging session can be understood based on historical session content and natural language processing. Session intentions may include consultation, after-sales, etc. In some embodiments, a knowledge base may be pre-established, and the knowledge base may classify multiple session intentions and keywords corresponding to various session intentions. Based on the historical session content of the other party of this instant messaging session, keywords can be extracted, and then the corresponding session intentions can be found in the knowledge base.

[0057] Inputting the conversation intent indicator as one of the prompt words into the preset large language model helps the preset large language model to convert the original text into the target text so that the generated target text is more in line with the semantic context of the current instant messaging session. When the interval between the historical conversation content and the current instant messaging session is short, the target text generated in combination with the conversation intent indicator helps to make the reply content of the user to the other party have a contextual association with the historical conversation content, and the experience on the other party side is that the customer service provided by the user side is coherent and non-fragmented.

[0058] Furthermore, the expression style may include concise expression, humorous expression, etc. The user may preset the expression style. For example, the expression style may be set according to the corporate image.

[0059] By inputting the expression style indicator as one of the prompt words into the preset large language model, the target text can be polished, the vividness and diversity of the reply content can be increased, and the adaptability of the reply content to the brand positioning can be improved.

[0060] For example, assuming the original text is "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities", and the prompt word includes the IP address (the corresponding second language is English), the target text output by the corresponding preset large language model is: Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities.

[0061] If the original text is the same, the prompt words include the IP address (corresponding to English as the second language) and the expression style indicator (indicating that the user's expression style is concise expression), the corresponding target text output by the preset large language model is: Sorry, no current discounts for this product. Follow for future promotions.

[0062] If the same original text, the prompt words include the IP address (corresponding to English as the second language) and the expression style indicator (indicating that the user's expression style is humorous expression), the corresponding preset large language model outputs the target text: Sorry, this product is on a "discount vacation" right now. But stay tuned! The discount party might just be around the corner.

[0063] Furthermore, the prompt word may also include a user persona indicator for indicating the user's persona. The user persona indicator may be preset by the user. For example, the user's persona may include industry information, such as beauty, maternal and child care, etc.).

[0064] Inputting the user persona indicator as one of the prompt words into the preset large language model can polish the target text so that the user's reply content is more in line with the industry / brand image.

[0065] For example, assuming the original text is "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities", and the prompt word includes the IP address (the corresponding second language is English), the target text output by the corresponding preset large language model is: Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities.

[0066] If the same original text, the prompt words include the IP address (corresponding to English as the second language) and the user persona indicator (indicating that the user's persona is a beauty customer service representative), the corresponding preset large language model outputs the target text: Dear, there is no discount for this product at the moment. But don't be disappointed! The following promotional activities are really worth waiting for. There might be big surprises. You must keep an eye on our store!

[0067] If the same original text, the prompt words include the IP address (corresponding to the second language is English) and the user's personality indicator (indicating that the user's personality is a light luxury saleswoman), the corresponding preset large language model outputs the target text: We'resorry, but there are no ongoing promotions for this item. However, we suggestyou keep an eye on our future activities. Perhaps the next exciting promotionwill allow you to acquire it at a more appealing price.

[0068] If the same original text, the prompt words include the IP address (corresponding to English as the second language) and the user persona indicator (indicating that the user's persona is a saleswoman in the jewelry industry), the corresponding target text output by the preset large language model is: Dear valued customer, we're truly sorry that this exquisite jewelry piece is not part of any current promotional activities. However, we sincerely invite you to keep a close eye on our upcoming promotional events. Each event is aromantic encounter with resplendent treasures, and we look forward to bringing you surprises beyond your imagination.

[0069] As a result, the target text output by the preset large language model can be more in line with what the user really wants to express, rather than a mechanical literal translation, and the final presentation is more in line with the language habits of the other party, improving communication accuracy and instant messaging experience. Specifically, in terms of translation capabilities, the terminal executing this implementation scheme can be similar to the role of intelligent customer service, relying on the powerful language understanding and generation capabilities of the preset large language model to smoothly convert the reply content written by the user in Chinese into the target language used by the other party. Through the application of prompt words, in the translation process, the preset large language model is not a simple word-to-word conversion, but fully considers the grammatical rules, semantic contexts, cultural backgrounds, and expression habits of different languages, and intelligently optimizes and polishes the translation, avoiding the common "stiff" and "Chinese translation" problems in traditional translation, thereby providing the other party with natural, authentic, and easy-to-understand translation results, greatly improving the communication efficiency and customer experience in cross-border communication scenarios, truly breaking down language barriers, and achieving seamless customer service interaction.

[0070] In a specific implementation, in a single instant messaging session with the other end, the prompt word may also include a temporary language indicator, which is determined according to the previous context of the current instant messaging session; wherein the second language is determined according to the temporary language indicator.

[0071] Specifically, the closing operation of the session interface can be used as the dividing point between the two instant messaging sessions. A single instant messaging session may last for a long time, and in the cross-border e-commerce scenario, the same account / device on the other end may be managed by different people. In this case, the actual user of the other end may change during a single instant messaging session. If the language terminals used by the actual users are different, the language of the other end may be temporarily switched on the session interface of the single instant messaging session.

[0072] Accordingly, the terminal can continuously monitor the conversation content of the current instant messaging conversation in the conversation interface, and if the language type of the latest conversation content is different from the language type of the previous content, it is determined that the other party has temporarily switched the language. Further, the temporary language is determined according to the language type of the latest conversation content and indicated by a temporary language indicator.

[0073] Further, the IP address and the temporary language indicator are both input as prompt words into the preset large language model, and the preset large language model can assign a third weight to the language type indicated by the temporary language indicator, and assign a second weight to the language type determined based on the IP address, and the third weight is higher than the second weight. In response to the language type indicated by the temporary language indicator being different from the language type determined based on the IP address, the preset large language model determines the language type indicated by the temporary language indicator with a higher weight as the target language (i.e., the second language) for this translation.

[0074] Alternatively, the IP address, temporary language indicator, and the language type with the highest usage rate in historical conversation content can all be input as prompt words into a preset large language model. The weights assigned by the preset large language model are sorted from high to low: third weight>first weight>second weight.

[0075] In some embodiments, when it is found through monitoring that the language types of the latest N conversation contents of the current instant messaging session are all different from the language types of the previous texts of the current instant messaging session, it is determined that a temporary language switch has occurred, and N is not less than 2. Thus, in the scenario where the language habit of the other party is to use multiple languages, it is possible to avoid erroneously triggering a temporary language switch and affecting the customer's communication experience.

[0076] In some embodiments, if the same temporary language indicator appears M times in a row (M is greater than or equal to 5), it means that the other party has changed a fixed person, and accordingly, the temporary language indicator can be determined as the default language of the other party. The terminal can record the number of consecutive appearances of the temporary language indicator. If the number of consecutive appearances reaches N times, the terminal directly determines the temporary language indicated by the temporary language indicator as the default language of the other party, and inputs the default language as the second language to the preset large language model. Accordingly, the preset large language model directly converts the original text into the target text expressed in the default language.

[0077] In some embodiments, as more conversation contents are sent using the temporary language, the temporary language may gradually accumulate to become the language type with the highest usage percentage in historical conversation contents, thereby affecting the second weight as a user feature.

[0078] Therefore, in a cross-border e-commerce scenario, the same account / device on the other end may be managed by different people logging in, and thus the language used by the other end may change in a single instant messaging session. This implementation scheme can automatically identify the change and generate a temporary language indicator. The preset large language model timely adjusts the second language used in the output target text according to the temporary language indicator, so that the second language used in the target text sent to the other end on the conversation interface can automatically change as the language used by the user on the other end changes.

[0079] In a specific implementation, in S104, the target text may be presented in a preview area of ​​the conversation interface; in response to receiving a confirmation instruction, the target text is presented in a conversation area of ​​the conversation interface; wherein the target text presented in the preview area is not visible to the peer end, and the target text presented in the conversation area is visible to the peer end.

[0080] In a typical application scenario, refer to Figure 2 , the other end or user can initiate an instant communication session with the other end, and accordingly, the display interface of the user side terminal and the display interface of the other end side network device both display the session interface 200. The session interface 200 may include a session area 204 for displaying the session content of the previous instant communication sessions between the two parties. Figure 2 The exemplary effect of the conversation interface 200 displayed on the display interface of the user-side terminal is exemplarily shown.

[0081] The conversation interface 200 may further include an input box 206 for allowing the user to input original text.

[0082] In a non-limiting example, after the user enters the original text in the input box 206, he can directly click the "Send" button. In response to the "Send" button being triggered, the terminal can directly send the original text to the other end. Specifically, the terminal can send the original text to the network device of the other end to present the original text in the conversation area 204 of the conversation interface 200 on the other end side (displayed through the display interface of the network device on the other end side), and the conversation interface 200 on the user side (displayed through the display interface of the user side terminal) can also synchronously display the original text sent this time. For example, if it is determined based on the historical conversation content with the other end that the common language of the other end is consistent with the common language of the user (that is, both are the first language), the original text can be sent directly to the other end.

[0083] Continue to refer Figure 2The conversation interface 200 may further include a preview area 202 , which is only displayed on the user side of the conversation interface 200 .

[0084] Specifically, the preview area 202 includes a target text display area and a function area. The function area may provide buttons such as "Translate", "Polish", "Cancel", and "Apply". The target text display area is used to display the target text output by the preset large language model for users to browse.

[0085] In a non-limiting example, the user enters the original text in the input box 206 and clicks the "Translate" button. In response to the "Translate" button being triggered, the terminal obtains the information of the other party (for example, the IP address of the other party), determines the prompt word based on the information of the other party, and inputs it into the preset large language model together with the original text, and displays the target text output by the preset large language model in the target text display area of ​​the preview area 202. For example, refer to Figure 2 , the original text entered in the input box 206 is "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities." The target text displayed in the target text display area is "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities." At this time, the target text is only presented on the session interface 200 displayed on the terminal side display interface, and the session interface 200 displayed on the display interface of the opposite side network device does not present the target text.

[0086] Further, the user clicks the "Apply" button to issue a confirmation instruction. In response to receiving the confirmation instruction, the terminal sends the target text to the network device on the other end to present the target text in the session area 204 of the session interface 200 on the other end (displayed through the display interface of the network device on the other end), and the session interface 200 on the user side (displayed through the display interface of the user side terminal) can also synchronously display the target text sent this time.

[0087] In some non-limiting examples, the conversation interface 200 on the user side may display the target text, such as Figure 2 In other non-limiting examples, the conversation interface 200 on the user side may display the original text, that is, the reply content with the same meaning is expressed and presented in different languages ​​on the user side and the other side, respectively, to facilitate understanding between both parties of the instant messaging conversation.

[0088] In a specific implementation, after the target text is presented in the preview area of ​​the conversation interface, the present embodiment may further include the steps of: receiving a first instruction, the first instruction being used to update a prompt word; inputting the original text and the updated prompt word into a preset large language model to obtain an updated target text; and presenting the updated target text in the preview area.

[0089] Continue to refer Figure 2 In response to the target text "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities." being presented in the preview area 202 of the conversation interface 200, the user can browse and check the content of the target text. If the user thinks that the current target text content does not meet expectations, the user can click the "polish" button to trigger the first instruction.

[0090] Specifically, the first instruction may be used to add or replace the prompt word determined in S102. For example, in response to the "polish" button being triggered, the terminal may pop up a prompt word selection box in the conversation interface 200, and the user may select the prompt word to be updated.

[0091] Assuming that the first instruction indicates that the newly added prompt word is an expression style indicator, and specifically indicates that the user's expression style is humorous expression, the terminal inputs the original text, the prompt word determined in S102, and the newly added expression style indicator into the preset large language model, and presents the updated target text output by the preset large language model in the target text display area of ​​the preview area 202, such as Figure 3 shown. Figure 3 In the input box 206, the original text entered is Figure 2 The examples shown are consistent, all of which are "Sorry, there is no discount for this product at the moment, please pay attention to subsequent promotions." Figure 3 In the example shown, the prompt words for inputting the preset large language model are added with expression style indicators, so Figure 3 The target text displayed in the target text display area is updated to "Sorry, this product is on a "discount vacation" right now. But staytuned! The discount party might just be around the corner."

[0092] Continue to refer Figure 3After the user browses and confirms that the updated target text meets expectations, he clicks the “Apply” button, and the terminal sends the updated target text to the other end, so that the updated target text is presented in the conversation area 204 .

[0093] In a specific implementation, after the target text is presented in the preview area of ​​the conversation interface, the method described in this embodiment may further include the steps of: receiving a second instruction, the second instruction is used to update the prompt word and indicate the target character, the target character is extracted from the target text; at least the original character corresponding to the target character and the updated prompt word are input into a preset large language model to obtain an updated target text, the original character is extracted from the original text; and presenting the updated target text in the preview area.

[0094] Continue to refer Figure 2 In response to the target text "Sorry, there is no discount for this product at the moment. Please keep an eye out for future promotional activities." being presented in the preview area 202 of the conversation interface 200, the user can browse and check the content of the target text. If the user is not satisfied with the translation of some words in the target text, the user can highlight these words and click the "polish" button to trigger the second instruction, and the highlighted words are the target characters.

[0095] In some embodiments, the terminal may input the original characters corresponding to the target characters in the original text and the updated prompt words into a preset large language model to obtain updated target characters; and replace the target characters with the updated target characters to obtain an updated target text.

[0096] In another embodiment, the terminal may input the original text and the updated prompt word into a preset large language model, the updated prompt word includes an update target indicator for indicating that the original character corresponding to the target character is to be translated, and the preset large language model directly outputs the updated target text, in which only the target character (and possibly the preceding and following characters that are grammatically associated with it) is updated.

[0097] In a specific implementation, after the target text is presented in the preview area of ​​the conversation interface, the method described in this embodiment may further include the step of: in response to receiving a third instruction, presenting a verification text in the preview area, the verification text being the target text expressed in the first language.

[0098] Specifically, the third instruction can be used to trigger a secondary translation check. For example, before the target text presented in the preview area 202 is sent to the peer, each click of the "Translate" button triggers a secondary translation check, which is used to switch the target text from the current language to the language before the most recent translation.

[0099] In response to receiving the third instruction, the terminal may input the target text and the verification indicator presented in the preview area into the preset large language model, and present the verification text output by the preset large language model in the preview area. The verification indicator is used to indicate the verification language, such as the first language. The preset large language model converts the target text into the verification text expressed in the verification language.

[0100] For example, refer to Figure 4 , the user enters the original text "Sorry, there is no discount for this product at the moment. You can pay attention to subsequent promotional activities." in the input box 206, and presets the user persona indicator as a beauty customer service. In response to the "Translate" button being triggered, the terminal inputs the original text, the IP address of the other end, and the user persona indicator into the preset large language model, and presents the target text "Dear, there is no discount for this product at the moment. Butdon't be disappointed! The following promotional activities are really worthwaiting for. There might be big surprises. You must keep an eye on ourstore!" output by the preset large language model in the preview area 202.

[0101] The user clicks the "Translate" button again to trigger the third instruction. The terminal inputs the verification indicator and the target text into the preset large language model, and the verification text "Baby, there is no discount for this product at the moment. But don't be discouraged. Subsequent promotions are worth waiting for. There may be a big surprise. You must continue to pay attention to us!" output by the preset large language model covers the target text and presents it in the preview area 202. Figure 5 shown.

[0102] Continue to refer Figure 5 After the user browses the verification text, he clicks the “Apply” button, and the terminal sends the target text to the other end and presents it in the conversation area 204 .

[0103] In some embodiments, the target characters in the second instruction can be extracted from the verification text. For example, when the user browses the verification text, he selects one or more words and clicks a "polish" button to trigger the second instruction.

[0104] In a typical application scenario, refer to Figure 6 , assuming that the second language determined based on the IP address of the other party is English, and in the current instant messaging session, the other party has interacted with the user in English at least once, but the latest consultation content sent by the other party is expressed in German, the terminal can determine that the language of the current instant messaging session with the other party has been temporarily switched.

[0105] Further, the user enters the original text "Sorry, there is no discount for this product at present, please pay attention to subsequent promotional activities" in Chinese (first language) in the input box 206 and clicks the "Translate" button.

[0106] The terminal inputs the original text and the temporary language indicator into the preset large language model, wherein the temporary language indicator indicates that the language type is German. The preset large language model sets the weight (i.e., the third weight) of the language type indicated by the temporary language indicator to the highest, that is, the weight of the language type indicated by the temporary language indicator is higher than the weight (second weight) of the language type determined based on the IP address of the other party in the current instant communication session. Further, the preset large language model determines the second language as German with a higher weight, and converts the original text into a target text expressed in German and presents it in the preview area 202.

[0107] If the user does not understand German, he can click the "Translate" button again to trigger the third instruction, and the terminal can call a third-party translation tool to translate the target text into a verification text expressed in the first language and present it in the preview area 202. The third-party translation tool can be any application / plug-in with translation function independent of the preset large language model.

[0108] After the user checks the text and confirms that it is correct, he clicks the "Apply" button to send the confirmation instruction. The terminal sends the target text to the other end, so that the target text expressed in the second language determined based on the temporary language indicator is presented in the conversation area 204, such as Figure 6 shown.

[0109] As a result, even if the user does not understand the second language at all, he can still achieve barrier-free instant communication with the other party.

[0110] In some embodiments, reference Figures 2 to 6 The preview area 202 may be superimposed on the session area 204 in the form of a layer.

[0111] In some embodiments, continue to refer to Figures 2 to 6 , the user can also click the "Cancel" button to delete the target text presented in the preview area 202. Thus, before the target text is sent to the other end, the user can cancel the translation operation at any time, thereby improving the flexibility of the cross-border e-commerce system.

[0112] From the above, this embodiment is used to automatically identify the language used by the other party (i.e., the second language) according to the IP address of the other party. As a result, the language used by the other party can be accurately identified at the first time when the instant messaging session is established, ensuring that the instant messaging session sent by the user to the other party is always presented in the language accustomed to by the other party, thereby improving the accuracy of communication. Furthermore, the language of the other party is preset according to the IP address, and after actually receiving the content sent by the other party (for example, consultation questions), the user can promptly feedback the reply content to the other party in the second language, thereby improving the efficiency of communication.

[0113] Furthermore, automatic translation of text is achieved by presetting a large language model to avoid situations such as stiff translation and Chinese-style translation, ensuring that the target text presented to the other party can accurately express the meaning of the original text expressed by the user in the first language and is easy for the other party to understand, thereby improving communication accuracy.

[0114] This implementation scheme can be applied to cross-border e-commerce scenarios, in which the counterpart can be an overseas customer using the cross-border e-commerce system, and the user can be a domestic customer service representative using the cross-border e-commerce system. The counterpart and the user implement cross-border e-commerce business through the cross-border e-commerce system. By accurately and automatically identifying the second language based on the IP address, and combining the preset large language model to accurately translate the original text input by the user into the target text, the customer service reply content is more in line with the customer's language habits and cultural background, thereby comprehensively improving the customer's consulting experience and helping domestic companies to steadily expand in overseas markets. For example, in a product consulting scenario, the counterpart can be an overseas consumer, and the user can be a customer service representative of the brand to which the product belongs. The use of this implementation scheme can enhance the consumer's brand pre-sales / after-sales experience.

[0115] Figure 7 Schematic diagram of the structure of an instant messaging device of a cross-border e-commerce system in the embodiment of the present disclosure. A person skilled in the art will appreciate that the instant messaging device 7 of the cross-border e-commerce system in the embodiment can be used to implement the above Figures 1 to 6 The method and technical solution described in the embodiments.

[0116] Specifically, refer to Figure 7The instant messaging device 7 of the cross-border e-commerce system may include: a receiving module 71, used to receive an original text that a user intends to send to a peer end, wherein the original text is expressed in a first language; a processing module 72, used to determine a prompt word according to information of the peer end, wherein, in the first instant messaging session with the peer end, the prompt word at least includes the IP address of the peer end; a generating module 73, used to input the original text and the prompt word into a preset large language model to obtain a target text, wherein the preset large language model is used to translate the original text into a target text according to the prompt word, wherein the target text is expressed in a second language, and the second language is determined at least according to the prompt word; a feedback module 74, used to present the target text on a conversation interface with the peer end, and send the target text to the peer end in response to a confirmation instruction of the user.

[0117] For more information about the working principle and working method of the instant messaging device 7 of the cross-border e-commerce system, please refer to the above Figures 1 to 6 The relevant description in the embodiment will not be repeated here.

[0118] The present disclosure also provides a computer-readable storage medium, which is a non-volatile storage medium or a non-transient storage medium, on which a computer program is stored. The computer program is executed by a processor. Figure 1 The steps of the instant messaging method of the cross-border e-commerce system provided in the illustrated embodiment. Preferably, the storage medium may include a computer-readable storage medium such as a non-volatile memory or a non-transitory memory. The storage medium may include a ROM, a RAM, a magnetic disk or an optical disk, etc.

[0119] The present disclosure also provides a terminal, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the above-mentioned Figure 1 The steps of the instant messaging method of the cross-border e-commerce system provided in the corresponding embodiment. The terminal may be a device deployed with the cross-border e-commerce system. The device may be, for example, a server, a mobile terminal, etc.

[0120] The present disclosure also provides a computer program product, including a computer program / instruction, which implements the above-mentioned Figure 1 The steps of the instant messaging method of the cross-border e-commerce system provided in the corresponding embodiment.

[0121] Although the disclosure is disclosed as above, the disclosure is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the disclosure, so the protection scope of the disclosure should be based on the scope defined by the claims.

Claims

1. An instant communication method for a cross-border e-commerce system, characterized in that: include: receiving an original text that a user intends to send to a peer end, wherein the original text is expressed in a first language; Determining a prompt word according to the information of the peer end, wherein, in the first instant messaging session with the peer end, the prompt word at least includes the IP address of the peer end; Inputting the original text and the prompt word into a preset large language model to obtain a target text, wherein the preset large language model is used to translate the original text into a target text according to the prompt word, wherein the target text is expressed in a second language, and the second language is determined at least according to the prompt word; The target text is presented on a conversation interface with the peer end, and in response to a confirmation instruction of the user, the target text is sent to the peer end.

2. The instant messaging method according to claim 1, characterized in that: In a non-first instant messaging session with the peer, the prompt words also include: User features of the peer end, wherein the user features are extracted from historical conversation content of the peer end; The second language is determined according to the user characteristics.

3. The instant messaging method according to claim 1, characterized in that: The prompt words also include: A language habit indicator, used to indicate the language habit of the other party; A session intention indicator, used to indicate the session intention of the peer end; An expression style indicator, used to indicate the expression style of the user; User personality indicator, used to indicate the user's personality.

4. The instant messaging method according to claim 1, characterized in that: In a single instant messaging session with the peer, the prompt words also include: A temporary language indicator, wherein the temporary language indicator is determined according to the previous context of the current instant messaging session; The second language is determined according to the temporary language indicator.

5. The instant messaging method according to claim 1, characterized in that: Presenting the target text on the conversation interface with the peer end includes: presenting the target text in a preview area of ​​the conversation interface; In response to the user's confirmation instruction, sending the target text to the peer terminal includes: in response to receiving the confirmation instruction, presenting the target text in the conversation area of ​​the conversation interface; The target text presented in the preview area is invisible to the peer end, and the target text presented in the conversation area is visible to the peer end.

6. The instant messaging method according to claim 5, characterized in that: After the target text is presented in the preview area of ​​the conversation interface, the method further includes: receiving a first instruction, wherein the first instruction is used to update the prompt word; Inputting the original text and the updated prompt words into the preset large language model to obtain an updated target text; The updated target text is presented in the preview area.

7. The instant messaging method according to claim 5, characterized in that: After the target text is presented in the preview area of ​​the conversation interface, the method further includes: receiving a second instruction, wherein the second instruction is used to update the prompt word and indicate a target character, wherein the target character is extracted from the target text; At least the original characters corresponding to the target characters and the updated prompt words are input into the preset large language model to obtain an updated target text, wherein the original characters are extracted from the original text; The updated target text is presented in the preview area.

8. The instant messaging method according to claim 5, characterized in that: After the target text is presented in the preview area of ​​the conversation interface, the method further includes: In response to receiving the third instruction, a verification text is presented in the preview area, where the verification text is the target text expressed in the first language.

9. An instant messaging device for a cross-border e-commerce system, characterized in that: include: A receiving module, used for receiving an original text that a user intends to send to a peer end, wherein the original text is expressed in a first language; A processing module, configured to determine a prompt word according to the information of the opposite end, wherein, in the first instant messaging session with the opposite end, the prompt word at least includes the IP address of the opposite end; a generating module, configured to input the original text and the prompt word into a preset large language model to obtain a target text, wherein the preset large language model is configured to translate the original text into a target text according to the prompt word, wherein the target text is expressed in a second language, and the second language is determined at least according to the prompt word; A feedback module is used to present the target text on a conversation interface with the opposite end, and send the target text to the opposite end in response to a confirmation instruction from the user.

10. A computer-readable storage medium, wherein the computer-readable storage medium is a non-volatile storage medium or a non-transient storage medium, and a computer program is stored thereon, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are performed.

11. A terminal comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor runs the computer program, the steps of the method according to any one of claims 1 to 8 are performed.

12. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images