Cross-border transaction information processing system and method based on large model

Abstract

The invention relates to a cross-border transaction information processing system and method based on a large model. The system comprises a tax collection module which is used for obtaining selling currency information, buying currency information and selling currency balance information; the information request module is used for inputting the selling currency information, the buying currency information and the selling currency balance information into a question generation model for statement generation processing to obtain request information, and if the request information meets a preset condition, the request information is determined as target request information; the data query module is used for vectorizing the target request information to obtain a target embedded vector, and querying a vector database based on the target embedded vector to obtain a target matching result; and the information generation module is used for inputting the target matching result into a preset large language model for text generation processing to obtain target output information corresponding to the target request information. According to the method and the device, professional and deep related information can be actively provided during exchange operation, the accuracy of user operation is improved, and the experience feeling of the user is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of information processing technology, and in particular to a cross-border transaction information processing system, method, storage medium and electronic device based on a large model. Background Art

[0002] With the growth of global cross-border e-commerce business, the compliance and policy issues faced by foreign trade enterprises in the process of collecting payments are becoming increasingly complex. In particular, for ordinary foreign trade enterprise users, the high threshold of professional information makes it easy for them to generate compliance risks or loss of revenue in actual operations. At present, the market mainly relies on fuzzy search in the knowledge base to answer user questions, but due to the low matching degree and untimely information update, it is difficult to effectively solve the deep-seated needs of users. In particular, the deficiencies in platform operation, compliance judgment and decision-making support have become the pain points of foreign trade enterprises' operations, and they are unable to actively provide users with professional information. Summary of the invention

[0003] In order to solve at least one of the technical problems mentioned above, the present disclosure proposes a cross-border transaction information processing system, method, storage medium and electronic device based on a big model.

[0004] According to one aspect of the present disclosure, there is provided a cross-border transaction information processing system based on a large model, which includes a tax exchange module, an information request module, a data query module, a vector database, and an information generation module;

[0005] The tax remittance module is used to obtain personal currency information, which includes selling currency information, buying currency information, and selling currency balance information;

[0006] The information request module is used to input the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information, and if the request information meets the preset conditions, determine the request information as the target request information;

[0007] The data query module is used to vectorize the target request information to obtain a target embedding vector, query the vector database based on the target embedding vector, and obtain a target matching result. The vector database is obtained by performing entity alignment on the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorizing the alignment results.

[0008] The information generation module is used to input the target matching result into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

[0009] In some possible implementations, the information request module includes:

[0010] A question generating unit, configured to input the selling currency information, the buying currency information and the selling currency balance information into a question generating model to obtain request information;

[0011] A concurrent control unit, configured to detect whether there is a token in the access token bucket, and if so, assign a first target token to the request information and delete the first target token from the access token bucket to obtain request information assigned with the first target token, wherein the first target token is any token in the access token bucket;

[0012] The target request information determining unit is used to determine whether the request information assigned to the first target token meets the preset condition, and determine the request information assigned to the first target token as the target request information.

[0013] In some possible implementations, the information request module further includes a question acquisition unit;

[0014] The question acquisition unit is used to acquire the request question input by the user;

[0015] The concurrent control unit is further used to detect whether there is a token in the access token bucket, and if so, assign a second target token to the request question and delete the second target token from the access token bucket to obtain a request question assigned with a second target token, wherein the second target token is any token in the access token bucket;

[0016] The target request information determination unit is further configured to determine that the request question assigned to the second target token meets the preset condition, and determine the request question assigned to the second target token as the target request information.

[0017] In some possible implementations, the system further includes a database updating module, and the database updating module includes:

[0018] A data set acquisition unit, used for periodically acquiring a target data set, wherein the target data set includes the question-and-answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set, and the foreign exchange information data set;

[0019] A graph construction unit, for constructing a first knowledge graph corresponding to the question-answer dataset, a second knowledge graph corresponding to the foreign trade foreign exchange basic dataset, a third knowledge graph corresponding to the foreign exchange policy dataset, and a fourth knowledge graph corresponding to the foreign exchange information dataset;

[0020] an alignment unit, configured to perform entity alignment on entities in the first knowledge graph, the second knowledge graph, the third knowledge graph, and the fourth knowledge graph to obtain the alignment result;

[0021] The data storage unit is used to vectorize the alignment result and store the obtained vectorized result in the vector database.

[0022] In some possible implementations, the system further includes a large language model updating module, and the large language model updating module includes:

[0023] An initial data set acquisition unit, used for periodically acquiring an initial data set, wherein the initial data set is obtained by preprocessing, aligning and normalizing the original data acquired from the question and answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set and the foreign exchange information data set;

[0024] A training data set determination unit, used to label the initial data set with question types to obtain a training data set;

[0025] The large language model determination unit is used to input the training data set into the initial multimodal large language model for supervised training to obtain the preset large language model.

[0026] In some possible implementations, the data query module includes:

[0027] A similarity determination unit, configured to perform vector similarity calculation processing on the target embedding vector and each record vector in the vector database to obtain at least one similarity;

[0028] The matching result determination unit is used to determine the record vector corresponding to the target similarity as the target matching result, wherein the target similarity is the similarity corresponding to the maximum value of the at least one similarity.

[0029] In some possible implementations, it is characterized in that

[0030] The tax collection module is also used to display the target output information.

[0031] According to a second aspect of the present disclosure, a method for processing cross-border transaction information based on a large model is provided, the method comprising:

[0032] Obtain personal currency information, including selling currency information, buying currency information, and selling currency balance information;

[0033] Inputting the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information, and if the request information meets a preset condition, determining the request information as target request information;

[0034] Vectorize the target request information to obtain a target embedding vector, query the vector database based on the target embedding vector to obtain a target matching result, wherein the vector database is obtained by entity alignment of the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorize the alignment results;

[0035] The target matching result is input into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

[0036] According to a third aspect of the present disclosure, there is provided an electronic device comprising at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor implements the large model-based cross-border transaction information processing method as described in any one of the first aspects by executing the instructions stored in the memory.

[0037] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the large model-based cross-border transaction information processing method as described in any one of the first aspects.

[0038] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

[0039] The implementation of this disclosure has the following beneficial effects:

[0040] The tax exchange module is used to obtain personal currency information, which includes selling currency information, buying currency information, and selling currency balance information; the information request module is used to input the selling currency information, buying currency information, and selling currency balance information into the question generation model for sentence generation processing to obtain request information. If the request information meets the preset conditions, the request information is determined as the target request information; the data query module is used to vectorize the target request information to obtain the target embedding vector, and query the vector database based on the target embedding vector to obtain the target matching result. The vector database is based on the entity alignment of the knowledge graphs corresponding to the question and answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set, and the foreign exchange information data set, and the alignment results are vectorized; the information generation module is used to input the target matching result into the preset large language model for text generation processing to obtain the target output information corresponding to the target request information. The vector database is based on the entity alignment of the knowledge graphs corresponding to the question and answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set, and the foreign exchange information data set, and the alignment results are vectorized. The information request module of the large-model cross-border transaction information processing system uses the user's personal currency information to input the question generation model to generate personalized question requests. The data query module queries the vector database for related foreign trade foreign exchange basic data, foreign exchange policy data and some information data to obtain personalized matching information. The information generation module inputs the personalized matching information into the large language model for sentence integration to obtain the final output information. It automatically provides relevant professional information based on the user's personal currency information, and can actively provide professional and in-depth relevant information during currency exchange operations, thereby improving the accuracy of user operations and enhancing the user experience.

[0041] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the embodiments or prior art descriptions. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 A schematic diagram showing the structure of a cross-border transaction information processing system based on a large model according to an embodiment of the present disclosure is shown;

[0044] Figure 2 A schematic diagram showing a flow chart of a cross-border transaction information processing method based on a large model according to an embodiment of the present disclosure;

[0045] Figure 3 A block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0046] Reference numerals

[0047] 10. Tax remittance module; 20. Information request module; 30. Data query module; 40. Vector database; 50. Information generation module; DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0050] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0051] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0052] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.

[0053] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0054] Figure 1 A schematic diagram of the structure of a cross-border transaction information processing system based on a large model according to an embodiment of the present disclosure is shown, Figure 1 As shown, the above system includes:

[0055] It includes a tax collection module 10, an information request module 20, a data query module 30, a vector database 40 and an information generation module 50;

[0056] The tax exchange module 10 is used to obtain personal currency information, which includes selling currency information, buying currency information, and selling currency balance information;

[0057] The information request module 20 is used to input the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information. If the request information meets the preset conditions, the request information is determined as the target request information;

[0058] The data query module 30 is used to vectorize the target request information to obtain the target embedding vector, query the vector database 40 based on the target embedding vector, and obtain the target matching result. The vector database 40 is obtained by entity alignment of the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorizing the alignment results;

[0059] The information generation module 50 is used to input the target matching result into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

[0060] In addition to providing a variety of basic cross-border financial service modules, the cross-border transaction information processing system based on the large model also includes a tax exchange module 10, an information request module 20, a data query module 30, a vector database 40 and an information generation module 50. In addition to being used to query the account balance information of various currencies of individuals and transaction operations between currencies, the tax exchange module 10 can also be used to obtain the user's personal currency information before the user confirms the transaction operation. The personal currency information includes but is not limited to the selling currency information, the buying currency information, and the selling currency balance information, and sends the personal currency information to the information request module 20. The information request module 20 determines the target request information based on the selling currency information, the buying currency information, the selling currency balance information and the question generation model. The question generation model is obtained by training the initial language generation model based on the question data set in the question and answer data set. The target matching result includes but is not limited to the target answer that matches the target embedding vector, and the foreign trade foreign exchange basic information, foreign exchange policy information and foreign exchange information information that support the target answer. The preset large language model is obtained by supervised training of the initial multimodal large language model based on a training data set obtained from a question and answer data set, a foreign trade foreign exchange basic data set, a foreign exchange policy data set, and a foreign exchange information data set.

[0061] The question-and-answer dataset is used to represent a collection of question-and-answer data related to cross-border financial services. The question-and-answer dataset can be directly obtained from the original cross-border trading platform. The question-and-answer dataset includes a question dataset and an answer dataset corresponding to the question dataset. For example, the question-and-answer dataset includes a collection of question-and-answer data between users and customer service in the cross-border trading platform. The foreign trade foreign exchange basic dataset is used to represent basic knowledge data related to foreign exchange and foreign trade. For example, the foreign trade foreign exchange basic dataset includes but is not limited to the classification of foreign exchange and the basic concepts of foreign exchange transactions. The foreign exchange policy dataset is used to represent official policies and legal regulations related to foreign exchange and foreign trade. For example, foreign exchange policy data includes but is not limited to the Regulations on the Management of Foreign Exchange Income, the Regulations on the Management of Foreign Exchange Reserves, and the Regulations on Foreign Exchange Control Measures. The foreign exchange information dataset is used to represent news data related to foreign exchange and foreign trade within a preset time.

[0062] In some embodiments, the initial language generation model includes but is not limited to a recurrent neural network (RNN), a transformer network (Transformer) and a long short-term memory network (LSTM). The application platform of the cross-border transaction information processing system based on the large model includes but is not limited to a foreign trade payment platform between businesses (Business-to-Business, B2B) and an e-commerce payment platform between businesses (Business to Consumer, B2C).

[0063] In some embodiments, the tax exchange module 10 includes a currency account balance unit and an initiating exchange unit. After the user logs into the system, the currency account balance unit can display the user's currency balance information. For example, the currency balance information includes but is not limited to the available balance and frozen balance of the US dollar, the available balance and frozen balance of the euro, the available balance and frozen balance of the British pound, and the available balance and frozen balance of the RMB out of the case. The initiating exchange unit is used for the quick exchange unit and the entrusted exchange unit. For example, the quick exchange unit includes the selling currency information, the buying currency information and the delivery date. The selling currency information includes the selling currency type and the selling currency amount, and the buying currency information includes the buying currency type and the buying currency amount. After the user inputs the selling currency type, the selling currency amount and the buying currency type, the tax exchange module 10 sends the acquired selling currency information, buying currency information and selling currency balance information to the information request module 20, and the information request module 20 generates a request question, i.e., a target request information. For example, the selling currency type is US dollars, the buying currency type is euros, and the US dollar balance is $7000. The personal currency information including the selling currency type is US dollars, the buying currency type is euros, and the US dollar balance is $7000 is sent to the information request module 20. The information request module 20 generates a model for the selling currency type is US dollars, the buying currency type is euros, and the US dollar balance is $7000. The target request information is "I have $7000. From the perspective of asset management, evaluate whether to exchange currency and indicate risks." The target request information is input into the data query module 30, and the data query module 30 vectorizes the target request information to obtain a target embedding vector, and queries the vector database 40 based on the target embedding vector to obtain a matching target matching result. The target matching results are input into the preset large language model for sentence generation to obtain the target output information.

[0064] In the above technical solution, the information request module 20 of the large-model cross-border transaction information processing system utilizes the user's personal currency information to input the question generation model to generate personalized question requests, and queries the vector database 40 through the data query module 30 for related foreign trade foreign exchange basic data, foreign exchange policy data and some information data to obtain personalized matching information. The information generation module 50 inputs the personalized matching information into the large language model for sentence integration to obtain the final output information, and automatically provides relevant professional information based on the user's personal currency information. It can actively provide professional and in-depth relevant information during currency exchange operations, improve the accuracy of user operations, and enhance the user experience.

[0065] In some embodiments, the information request module 20 includes:

[0066] A question generation unit, used for inputting the selling currency information, the buying currency information and the selling currency balance information into the question generation model to obtain request information;

[0067] A concurrent control unit, used for detecting whether there is a token in the access token bucket, and if so, assigning a first target token to the request information and deleting the first target token from the access token bucket, to obtain request information assigned with the first target token, wherein the first target token is any token in the access token bucket;

[0068] The target request information determining unit is used to determine whether the request information assigned to the first target token meets a preset condition, and determine the request information assigned to the first target token as the target request information.

[0069] The preset conditions include the conditions for being assigned a token. The concurrent control unit of the information request module 20 introduces a distributed cluster and a gateway service, and the concurrent control unit is implemented based on a token bucket algorithm. After the question generation unit of the information request module 20 obtains the request information based on the selling currency information, the buying currency information, the selling currency balance information and the question generation model, it sends it to the concurrent control unit. After receiving the request information, the concurrent control unit needs to take out a token from the token bucket. If the token exists, the token is assigned to the request information, indicating that the request is allowed. If there is no token, the request is rejected. The capacity of the token bucket is limited, and a maximum of a preset number of tokens can be stored. Tokens exceeding the preset number of capacity will be discarded. The concurrent control unit is also used to add tokens to the token bucket based on a preset rate. Therefore, the token bucket allows a certain burst of traffic, but the request rate in the long term cannot exceed the preset rate.

[0070] In some embodiments, the concurrent control unit uses distributed token bucket current limiting, and can store the number of tokens currently remaining in the token bucket and the timestamp of the last token update in the token bucket through the remote dictionary server (Remote Dictionary Server, Redis). Atomic operations are performed through the scripting language (Lua) to ensure concurrency safety. The new number of tokens is calculated according to the preset rate and added to the bucket, but it will not exceed the maximum capacity of the bucket. If the number of tokens remaining in the bucket is sufficient to process the current request, the request is allowed and the corresponding token is subtracted; otherwise, the request is rejected.

[0071] The above technical solution allows burst traffic through a concurrent control unit and ensures that the long-term request rate meets the current limiting standard, thereby realizing distributed token bucket current limiting. It can provide an efficient current limiting solution in the high concurrency scenario of a cross-border transaction information processing system based on a large model.

[0072] In some embodiments, the information request module 20 further includes a question acquisition unit;

[0073] A question acquisition unit, used for acquiring a request question input by a user;

[0074] The concurrency control unit is also used to detect whether there is a token in the access token bucket. If so, the second target token is assigned to the request question and the second target token is deleted from the access token bucket to obtain the request question assigned with the second target token, where the second target token is any token in the access token bucket.

[0075] The target request information determination unit is further configured to determine whether the request question assigned to the second target token meets a preset condition, and determine the request question assigned to the second target token as the target request information.

[0076] The preset conditions include the conditions for being assigned a token. The question acquisition unit is used to receive the request question input by the user and distribute the token through the concurrent control unit. The target request information determination unit is used to determine the request question assigned with the token as the target request information.

[0077] In some embodiments, after a user logs in to a cross-border transaction information processing system based on a large model, when performing a transaction operation, if there is a problem, the user can enter the problem acquisition unit and input a request question. Exemplarily, the request question includes "From the perspective of asset management, is there a risk in buying euros?" The concurrent control unit determines whether there is a token, distributes the token to the request question, and determines the request question with the token as the target request information.

[0078] The above technical solution, the cross-border transaction information processing system based on the large model can not only automatically generate questions, i.e. request information, but also receive questions input by users, meet users' various consulting needs, and provide personalized services.

[0079] In some embodiments, the system further includes a database update module, the database update module including:

[0080] A data set acquisition unit, used for periodically acquiring a target data set, the target data set includes a question-and-answer data set, a foreign trade foreign exchange basic data set, a foreign exchange policy data set, and a foreign exchange information data set;

[0081] A graph construction unit, used to construct a first knowledge graph corresponding to the question and answer data set, a second knowledge graph corresponding to the foreign trade foreign exchange basic data set, a third knowledge graph corresponding to the foreign exchange policy data set, and a fourth knowledge graph corresponding to the foreign exchange information data set;

[0082] An alignment unit, used to perform entity alignment on entities in the first knowledge graph, the second knowledge graph, the third knowledge graph, and the fourth knowledge graph to obtain an alignment result;

[0083] The data storage unit is used to vectorize the alignment results and store the obtained vectorized results in the vector database 40.

[0084] The data set acquisition unit of the cross-border transaction information processing system based on the big model can periodically acquire the target data set. The target data set of the cross-border transaction information processing system based on the big model can be updated in real time. The target data set includes but is not limited to the question and answer data set, the foreign trade and foreign exchange basic data set, the foreign exchange policy data set and the foreign exchange information data set. Among them, the question and answer data set can be acquired from the original cross-border transaction platform, and the newly added question and answer data can be acquired periodically to update the question and answer data set. The foreign trade and foreign exchange basic data set is used to characterize the basic knowledge data related to foreign exchange and foreign trade, and the newly added basic knowledge data related to foreign trade and foreign exchange can be acquired periodically to update the foreign trade and foreign exchange basic data set. Similarly, the foreign exchange policy data set and the foreign exchange information data set will also periodically acquire and update the newly added data respectively. The method for acquiring new data can be to synchronize messages through a message queue (MQ). For example, after the question and answer data set of the trading platform is updated, the data of the changed questions will be broadcast through MQ messages, and the system will monitor the messages and regenerate the latest question and answer data set; the method for acquiring new data can also be acquired through a hypertext transfer protocol (http) request. For example, by regularly tracking popular professional websites through a scheduled task, the latest basic knowledge data is collected using http requests to generate the latest foreign trade foreign exchange basic data set. The method for acquiring new data also includes manual acquisition.

[0085] In some embodiments, the graph construction unit is specifically used to construct a first knowledge graph by obtaining the entities and the relationships between the entities in the question data set and the answer data set corresponding to the question data set in the question and answer data set, construct a second knowledge graph by obtaining the entities and the relationships between the entities in the foreign trade foreign exchange basic data set, construct a third knowledge graph by obtaining the entities and the relationships between the entities in the foreign exchange policy data set, and construct a fourth knowledge graph by obtaining the entities and the relationships between the entities in the foreign exchange information data set. Among them, the entity is used to characterize the characteristic keywords. Exemplarily, the entity of the question "I have $7000 in my hand. From the perspective of asset management, evaluate whether to exchange currency and prompt risks" includes but is not limited to "$7000", "asset management", "assessment", "exchange" and "risk". The same entities in the first knowledge graph, the second knowledge graph, the third knowledge graph and the fourth knowledge graph are aligned to obtain the alignment results, and the entity results are vectorized and stored in the vector database 40. Among them, the vectorization processing method includes but is not limited to the word embedding method.

[0086] The above technical solution periodically uses the target data set to update the vector database 40. By integrating the question-and-answer knowledge base of the trading platform, the foreign exchange policy data set, national policies and regulations, foreign exchange information and other multi-dimensional data, it provides more accurate and in-depth answers to users' compliance, operation and decision support needs in the cross-border collection process, breaks the limitations of traditional knowledge bases, and improves the accuracy of information matching.

[0087] In some embodiments, the system further includes a large language model updating module, the large language model updating module including:

[0088] An initial data set acquisition unit, used for periodically acquiring an initial data set, wherein the initial data set is obtained by preprocessing, aligning and normalizing the original data acquired from the question-and-answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set and the foreign exchange information data set;

[0089] A training data set determination unit, used for labeling the initial data set with question types to obtain a training data set;

[0090] The large language model determination unit is used to input the training data set into the initial multimodal large language model for supervised training to obtain a preset large language model.

[0091] The acquisition method of the original data and the newly added data in the question and answer data set, foreign trade foreign exchange basic data set, foreign exchange policy data set and foreign exchange information data set corresponding to the initial data set in the initial data set acquisition unit is the same as the acquisition method of the question and answer data set, foreign trade foreign exchange basic data set, foreign exchange policy data set and foreign exchange information data set in the target data set. The acquired question and answer data set, foreign trade foreign exchange basic data set, foreign exchange policy data set and foreign exchange information data set are preprocessed to clean up the noise data and standardize the data format to obtain the initial data set, which can be updated periodically. The initial data set update method is the same as the target data set update method.

[0092] In some embodiments, the training data set determination unit uses a labeling tool to label the initial data set with question types, constructs a training data set, and transmits the training data set to the large language model determination unit. The large language model determination unit inputs the training data set into the initial multimodal large language model for supervised training to obtain a preset large language model. Among them, the labeling tools include but are not limited to You Only Look Once (YOLO) and the text annotation tool (Tagtog). The initial multimodal large language model includes but is not limited to the InternVL1.5 multimodal large language model.

[0093] The above technical solution integrates multi-dimensional data such as the trading platform's question-and-answer knowledge base, foreign exchange policy data set, national policies and regulations, and foreign exchange information, and uses large models for natural language training to provide more accurate and in-depth answers to users' compliance, operational and decision-making support needs during cross-border payment collection.

[0094] In some embodiments, the data query module 30 includes:

[0095] A similarity determination unit, configured to perform vector similarity calculation processing on the target embedding vector and each record vector in the vector database 40 to obtain at least one similarity;

[0096] The matching result determination unit is used to determine the record vector corresponding to the target similarity as the target matching result, where the target similarity is the similarity corresponding to the maximum value of at least one similarity.

[0097] Each record vector in the vector database 40 is used to represent a vector corresponding to each alignment result obtained based on the question-answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set, and the foreign exchange information data set. The similarity determination unit calculates the similarity value between the target embedding vector and each record vector in the vector database 40 based on the similarity calculation method to obtain at least one similarity, and the matching result determination unit uses the record vector with the largest similarity between the target embedding vector and each record vector in the vector database 40 as the target matching result.

[0098] In some embodiments, the similarity calculation method includes but is not limited to a cosine similarity algorithm.

[0099] The above technical solution uses vector similarity to highly match the target embedding vector with the vector database 40 to obtain accurate matching results, thereby providing more accurate and in-depth answers to users' compliance, operation and decision support needs during cross-border payment collection.

[0100] In some embodiments, it is characterized in that

[0101] The tax collection module 10 is also used to display target output information.

[0102] The tax collection module 10 also includes an information display unit, which is used to display the generated target data information to provide reference information for users when making decisions and enhance the user experience.

[0103] See also Figure 2 According to a second aspect of the present disclosure, a method for processing cross-border transaction information based on a large model is provided, the method comprising:

[0104] S101. Obtain personal currency information, which includes selling currency information, buying currency information, and selling currency balance information;

[0105] S102, inputting the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information, and if the request information meets the preset conditions, determining the request information as the target request information;

[0106] S103, vectorizing the target request information to obtain a target embedding vector, querying the vector database 40 based on the target embedding vector to obtain a target matching result, wherein the vector database 40 is obtained by performing entity alignment on the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorizing the alignment results;

[0107] S104: Input the target matching result into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

[0108] In some embodiments, when a user enters the tax remittance module 10 of the cross-border transaction information processing system of the large model and inputs the selling currency information and the buying currency information, the user's personal currency information is input into the question generation model to automatically generate request information, and the personal currency information includes but is not limited to the selling currency information, the buying currency information, and the selling currency balance information. Take a token from the token bucket. If the token exists, the first target token is assigned to the request information, indicating that the request is allowed. If there is no token, the request is rejected. The request information assigned to the first target token meets the preset conditions, and the request information assigned to the first target token is determined as the target request information. The question generation model is obtained by training the initial language generation model based on the question data set in the question and answer data set. After the target request information is vectorized, a target embedding vector is obtained, and the target embedding vector is input into the vector database 40 for matching to obtain a target matching result. The target matching result is not limited to a target answer that matches the target embedding vector, as well as foreign trade foreign exchange basic information, foreign exchange policy information, and foreign exchange information information that support the target answer. The target matching result is input into the preset large language model for text generation to obtain the target output information. The preset large language model is obtained by supervised training of the initial multimodal large language model based on the training data set obtained from the question and answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set and the foreign exchange information data set.

[0109] In another embodiment, when a user enters the question acquisition unit of the cross-border transaction information processing system of the large model and inputs a request question, a token is taken out from the token bucket. If the token exists, the second target token is assigned to the request question, indicating that the request is allowed. If there is no token, the request is rejected. The request question assigned with the second target token is determined as the target request information. The processing method of the target request information is the same as the above processing method, which is not repeated here.

[0110] The above technical solution utilizes the user's personal currency information to input the question generation model, generates personalized question requests, queries the relevant foreign trade foreign exchange basic data, foreign exchange policy data and some information data through the database, obtains personalized matching information, inputs the personalized matching information into the large language model for sentence integration, and obtains the final output information. It automatically provides relevant professional information based on the user's personal currency information, and can actively provide professional and in-depth relevant information during currency exchange operations, thereby improving the accuracy of user operations and enhancing the user experience.

[0111] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0112] An embodiment of the present application provides a cross-border transaction information processing device based on a big model, which may be a terminal or a server. The cross-border transaction information processing device based on a big model includes a processor and a memory, in which at least one instruction or at least one program is stored. The at least one instruction or the at least one program is loaded and executed by the processor to implement the cross-border transaction information processing method based on a big model provided in the above method embodiment.

[0113] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.

[0114] The method embodiments provided in the embodiments of the present application can be executed in electronic devices such as mobile terminals, computer terminals, servers or similar computing devices. Figure 31 is a hardware structure block diagram of an electronic device for a cross-border transaction information processing method based on a large model provided in an embodiment of the present application. Figure 3 As shown, the electronic device 900 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 910 (the processor 910 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 930 for storing data, and one or more storage media 920 (such as one or more mass storage devices) for storing application programs 923 or data 922. Among them, the memory 930 and the storage medium 920 can be short-term storage or permanent storage. The program stored in the storage medium 920 may include one or more modules, each of which may include a series of instruction operations in the electronic device. Furthermore, the central processing unit 910 can be configured to communicate with the storage medium 920 to execute a series of instruction operations in the storage medium 920 on the electronic device 900. The electronic device 900 may also include one or more power supplies 960, one or more wired or wireless network interfaces 950, one or more input and output interfaces 940, and / or one or more operating systems 921, such as Windows ServerTM, Mac OSXTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0115] The input / output interface 940 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the electronic device 900. In one example, the input / output interface 940 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices via a base station so as to communicate with the Internet. In one example, the input / output interface 940 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0116] It can be understood by those skilled in the art that Figure 3 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 3 More or fewer components as shown, or with Figure 3 Different configurations are shown.

[0117] An embodiment of the present application also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to a cross-border transaction information processing method based on a big model in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the cross-border transaction information processing method based on a big model provided in the above method embodiment.

[0118] Optionally, in this embodiment, the storage medium may be located in at least one of the multiple network servers of the computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0119] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above-mentioned various optional implementations.

[0120] From the above embodiments of the cross-border transaction information processing system, method, device, terminal, server, storage medium or computer program based on the big model provided by the present application, it can be seen that the information request module of the cross-border transaction information processing system of the big model of the present application uses the user's personal currency information to input the question generation model to generate personalized question requests, and queries the foreign trade foreign exchange basic data, foreign exchange policy data and some information data related to the vector database through the data query module to obtain personalized matching information. The information generation module inputs the personalized matching information into the big language model for sentence integration to obtain the final output information, and automatically provides relevant professional information according to the user's personal currency information. It can actively provide professional and in-depth relevant information during the exchange operation, improve the accuracy of the user's operation, and enhance the user's experience. .

[0121] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0123] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or may be accomplished by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0124] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A cross-border transaction information processing system based on a large model, characterized in that: The system includes a tax collection module, an information request module, a data query module, a vector database and an information generation module; The tax remittance module is used to obtain personal currency information, which includes selling currency information, buying currency information, and selling currency balance information; The information request module is used to input the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information, and if the request information meets the preset conditions, determine the request information as the target request information; The data query module is used to vectorize the target request information to obtain a target embedding vector, query the vector database based on the target embedding vector, and obtain a target matching result. The vector database is obtained by performing entity alignment on the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorizing the alignment results. The information generation module is used to input the target matching result into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

2. The system according to claim 1, characterized in that The information request module comprises: A question generating unit, configured to input the selling currency information, the buying currency information and the selling currency balance information into a question generating model to obtain request information; A concurrent control unit, configured to detect whether there is a token in the access token bucket, and if so, assign a first target token to the request information and delete the first target token from the access token bucket to obtain request information assigned with the first target token, wherein the first target token is any token in the access token bucket; The target request information determining unit is used to determine whether the request information assigned to the first target token meets the preset condition, and determine the request information assigned to the first target token as the target request information.

3. The system according to claim 2, characterized in that The information request module also includes a question acquisition unit; The question acquisition unit is used to acquire the request question input by the user; The concurrent control unit is further used to detect whether there is a token in the access token bucket, and if so, assign a second target token to the request question and delete the second target token from the access token bucket to obtain a request question assigned with a second target token, wherein the second target token is any token in the access token bucket; The target request information determination unit is further configured to determine that the request question assigned to the second target token meets the preset condition, and determine the request question assigned to the second target token as the target request information.

4. The system according to claim 1, characterized in that The system also includes a database update module, which includes: A data set acquisition unit, used for periodically acquiring a target data set, wherein the target data set includes the question-and-answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set, and the foreign exchange information data set; A graph construction unit, for constructing a first knowledge graph corresponding to the question-answer dataset, a second knowledge graph corresponding to the foreign trade foreign exchange basic dataset, a third knowledge graph corresponding to the foreign exchange policy dataset, and a fourth knowledge graph corresponding to the foreign exchange information dataset; an alignment unit, configured to perform entity alignment on entities in the first knowledge graph, the second knowledge graph, the third knowledge graph, and the fourth knowledge graph to obtain the alignment result; The data storage unit is used to vectorize the alignment result and store the obtained vectorized result in the vector database.

5. The system according to claim 1, characterized in that The system further includes a large language model updating module, wherein the large language model updating module includes: An initial data set acquisition unit, used for periodically acquiring an initial data set, wherein the initial data set is obtained by preprocessing, aligning and normalizing the original data acquired from the question and answer data set, the foreign trade foreign exchange basic data set, the foreign exchange policy data set and the foreign exchange information data set; A training data set determination unit, used to label the initial data set with question types to obtain a training data set; The large language model determination unit is used to input the training data set into the initial multimodal large language model for supervised training to obtain the preset large language model.

6. The system according to claim 1, characterized in that The data query module includes: A similarity determination unit, configured to perform vector similarity calculation processing on the target embedding vector and each record vector in the vector database to obtain at least one similarity; The matching result determination unit is used to determine the record vector corresponding to the target similarity as the target matching result, wherein the target similarity is the similarity corresponding to the maximum value of the at least one similarity.

7. The system according to claim 1, characterized in that The tax collection module is also used to display the target output information.

8. A method for processing cross-border transaction information based on a large model, applied to the cross-border transaction information processing system based on a large model as claimed in any one of claims 1 to 7, characterized in that: The method comprises: Obtain personal currency information, including selling currency information, buying currency information, and selling currency balance information; Inputting the selling currency information, the buying currency information and the selling currency balance information into the question generation model for sentence generation processing to obtain request information, and if the request information meets a preset condition, determining the request information as target request information; Vectorize the target request information to obtain a target embedding vector, query the vector database based on the target embedding vector to obtain a target matching result, wherein the vector database is obtained by entity alignment of the knowledge graphs corresponding to the question and answer dataset, the foreign trade foreign exchange basic dataset, the foreign exchange policy dataset, and the foreign exchange information dataset, and vectorize the alignment results; The target matching result is input into a preset large language model for text generation processing to obtain target output information corresponding to the target request information.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the cross-border transaction information processing method based on a large model as described in claim 8.

10. An electronic device, characterized in that: It includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor implements the large model-based cross-border transaction information processing method as described in claim 8 by executing the instructions stored in the memory.

Images