Method and device for processing digital RMB related data based on large model technology

Abstract

The present application discloses a method and device for processing digital RMB-related data based on large model technology. Among them, the method includes: obtaining multimodal data involved in the transaction event flow of digital RMB, and mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data; extracting target triples from the target data, and constructing an entity relationship network based on the target triples to obtain a target knowledge graph; using a graph-enhanced reasoning model, combined with risk management rules, to identify abnormal paths in the target knowledge graph, wherein the abnormal paths correspond to abnormal processing behaviors of digital RMB, and risk management rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behaviors of digital RMB. The present application solves the technical problem that the data processing methods in the related technologies are difficult to meet the data mining and intelligent analysis requirements of the multi-dimensional semantic association density of digital RMB.

Description

Technical Field

[0001] This application relates to the field of financial data processing technology, and specifically, to a method and device for processing digital RMB-related data based on big model technology. Background Art

[0002] With the rapid development of the digital economy, the digital RMB has gradually entered the stage of practical application and plays an important role in multiple scenarios such as retail payments, cross-border payments, and the disbursement of fiscal subsidies. The data types involved are complex, including user behavior data, transaction data, payment channel data, contract execution data, etc., showing the coexistence of highly structured and unstructured data and rapid dynamic evolution.

[0003] Against this backdrop, efficient processing and intelligent analysis of digital RMB-related data has become a pressing issue. Currently, data processing methods in related technologies mostly rely on rule engines and manually configured analysis models. These methods suffer from slow response, weak semantic understanding, and low data integration, making them unable to meet the multi-dimensional, semantically intensive data mining and intelligent decision-making requirements of the digital RMB.

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] The embodiments of the present application provide a method and device for processing digital RMB-related data based on big model technology, so as to at least solve the technical problem that the data processing methods in related technologies are difficult to meet the data mining and intelligent analysis requirements of the multi-dimensional semantic association density of digital RMB.

[0006] According to one aspect of an embodiment of the present application, a method for processing digital RMB-related data based on big model technology is provided, including: obtaining multimodal data involved in the transaction event stream of digital RMB, and mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data, wherein, in the unified semantic space, the more semantically similar the data, the closer the distance between the vectors corresponding to the data, and the greater the semantic difference, the farther the distance between the vectors corresponding to the data; extracting target triples from the target data, and constructing an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to characterize the association relationship between two entities; using a graph-enhanced reasoning model, combined with risk management rules, to identify abnormal paths in the target knowledge graph, wherein the abnormal paths correspond to abnormal processing behaviors of digital RMB, and the risk management rules are used to indicate the characteristics and patterns of rules that should be followed for the processing behaviors of digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and big model technology.

[0007] Optionally, the multimodal data includes: structured data and unstructured data of different modalities, wherein the structured data includes at least one of the following: transaction records, account information, device binding records, and the unstructured data includes at least one of the following: policy text, user behavior logs, contract descriptions, and image credentials; mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data includes: using a modal encoder in a semantic alignment model to extract semantic features corresponding to data of different modalities, wherein the feature extraction methods corresponding to data of different modalities are different, and the modalities include at least one of the following: text, voice, image, and video; using a connector in a semantic alignment model to convert the semantic features corresponding to different modalities, wherein the conversion processing is used to eliminate the differences in distribution, scale, and noise of semantic features of different modalities, and obtain semantic features in a unified representation form to ensure that data of all modalities can be represented in the same semantic space; using a generator in a semantic alignment model to map the converted semantic features to a unified semantic space to obtain target data.

[0008] Optionally, the training steps of the semantic alignment model include: obtaining a sample data set, wherein the sample data set contains multiple sample data, and each sample data contains data of different modalities for describing the same content; performing feature extraction on the data of different modalities in the sample data to obtain feature vectors corresponding to different modalities; pairing the feature vectors of different modalities corresponding to the same sample data to obtain positive sample pairs, and randomly selecting feature vectors of different modalities corresponding to different sample data to pair to obtain negative sample pairs; forming a training data set with positive samples and negative sample pairs, and based on the training data set, training the initial model in combination with the contrast loss function to obtain a semantic alignment model, wherein the contrast loss function is used to make the feature vectors in the positive sample pairs in the results output by the initial model closer in the semantic space, while the feature vectors in the negative sample pairs are farther apart in the semantic space during the training process.

[0009] Optionally, the method also includes: determining the data source of the multimodal data; when the data source represents that the text data in the multimodal data comes from different financial institutions, extracting target keywords from the text data of different financial institutions, wherein the target keywords include at least one of the following: terminology expressions, specific codes, and behavior descriptions within each financial institution; according to the standard expression mapping table, converting the keywords corresponding to different financial institutions into a unified standard vocabulary, wherein the standard expression mapping table is a mapping table containing special terms of different financial institutions to a unified standard vocabulary, and the special terms of different financial institutions include the data expressions, feature codes, and behavior descriptions of each financial institution.

[0010] Optionally, constructing an entity relationship network based on the target triples to obtain a target knowledge graph includes: using a stream processing engine and message middleware to monitor the transaction event stream, and when the multimodal data corresponding to the transaction event stream is updated, extracting the latest target triples from the target data corresponding to the updated multimodal data; based on the latest target triples, performing entity change perception and relationship evolution detection to obtain change information corresponding to the constructed target knowledge graph, wherein entity change perception and relationship evolution detection are used to determine the changed entity elements and / or relationship elements in the latest target triples compared to the constructed target knowledge graph; based on the change information, performing an update operation on the constructed target knowledge graph to obtain the latest target knowledge graph, wherein the update operation includes: node insertion, edge creation, and attribute update.

[0011] Optionally, using a graph-enhanced reasoning model in combination with risk management rules to identify abnormal paths in a target knowledge graph includes: in a graph neural network of the graph-enhanced reasoning model, using a neighbor aggregation mechanism and a relationship perception mechanism to analyze feature information of entity nodes and edges between entity nodes in the target knowledge graph, and generating a node representation corresponding to each entity node containing contextual information, wherein the neighbor aggregation mechanism is used to integrate feature information of entity nodes and neighbor nodes corresponding to the entity nodes, and the relationship perception mechanism is used to model the impact of different types of edges on information transmission, the feature information of the entity nodes includes at least one of the following: account attributes, transaction behavior characteristics, and contract status information, and the feature information of the edges includes at least one of the following: transaction amount, time interval, and call relationship; based on the node representation corresponding to each entity node in the target knowledge graph, identifying the behavior pattern corresponding to each node path in the target knowledge graph, wherein the behavior pattern includes at least one of the following: account fund jump, multi-contract call chain, and multi-account association control; using a large language model in the graph-enhanced reasoning model, based on the risk management rules corresponding to each behavior pattern, performing logical constraint verification and conflict detection on the node path, obtaining the risk probability corresponding to each node path, and determining the node path whose risk probability exceeds the preset risk probability as an abnormal path.

[0012] Optionally, the method also includes: when an update of the policy text related to the digital RMB is detected in the information source, obtaining the latest policy text; using a large language model to analyze the policy text to identify the constraint entities in the policy text, as well as the restrictions and business logic corresponding to the constraint entities; based on the restrictions and business logic, generating a grammatical constraint graph structure corresponding to the constraint entities, and converting the grammatical constraint graph structure into corresponding structured risk management rules.

[0013] According to another aspect of the embodiment of the present application, a processing device for digital RMB related data based on big model technology is also provided, including: a semantic alignment module, used to obtain multimodal data involved in the transaction event stream of digital RMB, and map the multimodal data to a unified semantic space for semantic alignment to obtain target data, wherein, in the unified semantic space, the more similar the semantics, the closer the distance between the vectors corresponding to the data, and the greater the semantic difference, the farther the distance between the vectors corresponding to the data; a network construction module, used to extract target triples from the target data, and construct an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to characterize the association relationship between two entities; a risk identification module, used to use a graph-enhanced reasoning model, combined with risk control rules, to identify abnormal paths in the target knowledge graph, wherein the abnormal paths correspond to abnormal processing behaviors of digital RMB, and the risk control rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behaviors of digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and big model technology.

[0014] According to another aspect of the embodiments of the present application, an electronic device is also provided, including: a memory and a processor, the processor being used to run a program stored in the memory, wherein the program executes a method for processing digital RMB related data based on large model technology when running.

[0015] According to another aspect of an embodiment of the present application, a non-volatile storage medium is also provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes a method for processing digital RMB-related data based on big model technology by running the computer program.

[0016] In an embodiment of the present application, multimodal data involved in the transaction event stream of the digital RMB is obtained, and the multimodal data is mapped to a unified semantic space for semantic alignment to obtain target data, wherein, in the unified semantic space, the more similar the semantics of the data, the closer the distance between the vectors corresponding to the data, and the greater the semantic difference, the farther the distance between the vectors corresponding to the data; target triples are extracted from the target data, and an entity relationship network is constructed based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to characterize the association relationship between the two entities; a graph-enhanced reasoning model is used, combined with risk management rules, to identify abnormal paths in the target knowledge graph, wherein, The abnormal path corresponds to the abnormal processing behavior of the digital RMB. The risk management rules are used to indicate the characteristics and patterns of the rules that should be followed in the processing behavior of the digital RMB. The graph-enhanced reasoning model is a reasoning framework that combines graph neural networks and big model technologies. By introducing cross-modal embedding models and a unified semantic space construction mechanism, semantic alignment and knowledge extraction of multi-source data are performed. By building a big model-driven semantic-aware risk control engine, context modeling and intention reasoning of transaction behaviors are realized, achieving the purpose of effectively mining and analyzing digital RMB related data, and thus solving the technical problem that the data processing methods in related technologies are difficult to meet the data mining and intelligent analysis needs of the multi-dimensional semantic association intensiveness of the digital RMB. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 It is a hardware structure block diagram of a computer terminal (or electronic device) for implementing a method for processing digital RMB-related data based on large model technology according to an embodiment of the present application;

[0019] Figure 2 It is a schematic diagram of a method flow for processing digital RMB related data based on large model technology provided in an embodiment of the present application;

[0020] Figure 3 This is a schematic diagram of a risk identification evolution modeling for the financial field provided according to an embodiment of the present application;

[0021] Figure 4 This is a structural diagram of a device for processing digital RMB-related data based on big model technology provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0024] Among related technologies, risk monitoring and data processing technologies for digital RMB still have structural technical shortcomings in key dimensions such as intelligence, real-time performance, semantic understanding, multimodal integration, and system scalability, making it difficult to meet the regulatory and operational requirements of financial institutions for compliance, security, and high availability. Specifically,

[0025] 1) Semantic-level fusion and unified modeling of multi-source heterogeneous financial data;

[0026] The data involved in the digital RMB ecosystem possesses significant "4V" characteristics: volume, velocity, variety, and veracity. Data types range from structured transaction records, account information, and device identifiers to unstructured multimodal data sources such as policy texts, user behavior logs, contract descriptions, and image credentials. Furthermore, due to inconsistent data encoding standards across financial institutions and platforms, data structures exhibit strong heterogeneity. Traditional data integration solutions based on the digital RMB suffer from low information extraction accuracy, poor semantic alignment, and high information loss during the integration process, severely restricting the unified expression and in-depth analysis of data value.

[0027] 2) Dynamic identification and real-time response to new financial risk patterns;

[0028] The programmability and decentralized nature of the digital RMB make it difficult for traditional risk control methods, which rely on static rule bases, to capture potentially complex and risky behaviors. According to the latest monitoring data, the evolution cycle of new attack vectors has been shortened from 36 months in traditional finance to 715 days. Complex risk paths such as cross-chain transactions and contract jumps have increased to 38%, and zero-day attacks account for 21% of all high-risk behaviors. Existing batch-based monitoring systems generally have a response delay of 4 to 6 hours, resulting in approximately 19% of high-risk transactions going undetected.

[0029] 3) The digital RMB policy system exhibits characteristics of frequent updates, deep structural hierarchies, and complex semantic constraints. Regulatory policies are updated an average of 23 times per month and are sourced from multiple authoritative sources. Their textual content is semantically complex, with each policy containing an average of 57 implicit business logic constraints. Traditional rule modeling methods based on manual transformation are extremely inefficient.

[0030] In order to solve the above problems, relevant solutions are provided in the embodiments of the present application, which are described in detail below.

[0031] According to an embodiment of the present application, an embodiment of a method for processing digital RMB related data based on big model technology is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0032] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The following is a hardware structure diagram of a computer terminal (or electronic device) for implementing a method for processing digital RMB related data based on large model technology. Figure 1 As shown, the computer terminal 10 (or electronic device) may include one or more processors 102 (illustrated as 102a, 102b, ..., 102n in the figure) (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1More or fewer components than shown, or with Figure 1 Different configurations shown.

[0033] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be fully or partially integrated into any of the other components of the computer terminal 10 (or electronic device). As discussed in the embodiments of this application, the data processing circuitry serves as a processor control (e.g., selecting a variable resistor terminal path connected to an interface).

[0034] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for processing digital RMB related data based on large model technology in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned method for processing digital RMB related data based on large model technology. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include but are not limited to the Internet, corporate intranet, local area network, mobile communication network and a combination thereof.

[0035] Transmission device 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of computer terminal 10. In one embodiment, transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 106 may be a radio frequency (RF) module configured to communicate with the Internet wirelessly.

[0036] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or electronic device).

[0037] Under the above operating environment, the embodiment of this application provides a method for processing digital RMB related data based on large model technology. Figure 2This is a schematic diagram of a method flow for processing digital RMB related data based on large model technology according to an embodiment of the present application, such as Figure 2 As shown, the method includes the following steps:

[0038] Step S202: Acquire the multimodal data involved in the transaction event stream of the digital RMB, and map the multimodal data into a unified semantic space for semantic alignment to obtain target data. In the unified semantic space, the vectors corresponding to data with more similar semantics are closer, and the vectors corresponding to data with greater semantic differences are farther apart.

[0039] Step S204: extract target triples from the target data, and construct an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to represent the association relationship between the two entities;

[0040] Step S206: Use a graph-enhanced reasoning model, combined with risk management rules, to identify abnormal paths in the target knowledge graph, where the abnormal paths correspond to abnormal processing behaviors of the digital RMB. The risk management rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behaviors of the digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and large model technologies.

[0041] Through the above steps, by introducing a cross-modal embedding model and a unified semantic space construction mechanism, semantic alignment and knowledge extraction of multi-source data are carried out, and by building a semantic-aware risk control engine driven by a large model, context modeling and intention reasoning of transaction behaviors are realized, thereby achieving the purpose of effectively mining and analyzing digital RMB related data, and thus solving the technical problem that the data processing methods in related technologies are difficult to meet the data mining and intelligent analysis needs of the digital RMB's multi-dimensional semantic association density.

[0042] The following further introduces the method for processing digital RMB related data based on big model technology in steps S202 to S206 of the embodiment of this application.

[0043] The embodiments of this application focus on the issues of data diversity and semantic complexity in the digital RMB ecosystem, and propose a new multimodal semantic governance system with cross-structure, multimodal, and cross-system semantic connectivity and knowledge extraction capabilities, which are described in detail below.

[0044] First, the multimodal data involved in the digital RMB transaction event stream is acquired. For example, a highly adaptable data access layer can be used to collect structured data (such as account information, fund details, and device binding records) and unstructured data (such as central bank policy documents, scenario agreement contract texts, transaction voucher images, and device behavior logs) from financial institutions' core systems, regulatory platforms, and smart contract execution platforms. This data is collected in near real time through a standardized data bus model and asynchronous message queue processing mechanism, providing high-throughput and high-reliability collection capabilities.

[0045] Then, cross-modal representation learning can be performed on the collected multimodal data related to the digital RMB. By introducing a unified embedding space of financial semantics, data of different modalities can be made comparable and vector aligned in the semantic dimension. The specific steps are as follows.

[0046] In some embodiments of the present application, multimodal data includes: structured data and unstructured data of different modalities, wherein the structured data includes at least one of the following: transaction records, account information, and device binding records; the unstructured data includes at least one of the following: policy text, user behavior logs, contract descriptions, and image credentials; mapping the multimodal data into a unified semantic space for semantic alignment to obtain target data includes the following steps: using a modal encoder in a semantic alignment model to extract semantic features corresponding to data of different modalities, wherein the feature extraction methods corresponding to data of different modalities are different, and the modalities include at least one of the following: text, voice, image, and video; using a connector in a semantic alignment model to convert the semantic features corresponding to different modalities, wherein the conversion processing is used to eliminate the differences in distribution, scale, and noise of semantic features of different modalities, and obtain semantic features in a unified representation form to ensure that data of all modalities can be represented in the same semantic space; using a generator in a semantic alignment model to map the converted semantic features into a unified semantic space to obtain target data.

[0047] Specifically, the Modality Encoder of the semantic alignment model can be used to convert raw data of different modalities (images, audio, video, and text) into feature vectors (semantic features). Data of each modality can then be converted into a format that can be processed by the model using a specific encoder. For example, images and videos may be encoded using convolutional neural networks, while text may be encoded using language models such as word embeddings. After the feature vectors (semantic features) are extracted, these feature vectors of different modalities can be converted into a unified representation using a connector to ensure that data of all modalities can be represented in the same semantic space. Finally, the generator is responsible for producing the final output, mapping the converted semantic features into a unified semantic space to obtain the target data.

[0048] For example, using three types of heterogeneous modal data—text, images, and behavioral logs—a method combining cross-modal representation learning and contrastive learning can be used to achieve semantic consistency modeling and vector space alignment. Specifically, text data is extracted using pre-trained language models (such as BERT and RoBERTa), image data is encoded into visual feature vectors using a visual Transformer, and behavioral log data is extracted using a sequence modeling network (such as Transformer-XL). To ensure semantic consistency among different modal data in a unified space, the system introduces a multimodal contrastive loss function based on the comparison of positive and negative samples. This encourages text, image, and log vectors originating from the same financial entity or behavioral event to be close in space, while samples with different semantic meanings remain distant. Furthermore, modality-specific normalization strategies and scale alignment techniques are employed to quantify differences in distribution, scale, and noise level among features from different modalities, ensuring consistency and stability in the semantic space.

[0049] The training process of the above semantic alignment model is introduced below.

[0050] In some embodiments of the present application, the training steps of the semantic alignment model include: obtaining a sample data set, wherein the sample data set contains multiple sample data, and each sample data contains data of different modalities for describing the same content; performing feature extraction on the data of different modalities in the sample data to obtain feature vectors corresponding to different modalities; pairing the feature vectors of different modalities corresponding to the same sample data to obtain positive sample pairs, and randomly selecting feature vectors of different modalities corresponding to different sample data to pair to obtain negative sample pairs; forming a training data set with positive samples and negative sample pairs, and training the initial model based on the training data set in combination with the contrast loss function to obtain a semantic alignment model, wherein the contrast loss function is used to make the feature vectors in the positive sample pairs in the results output by the initial model closer in the semantic space, while the feature vectors in the negative sample pairs are farther apart in the semantic space during the training process.

[0051] For example, the training method is explained using data from two modalities, text and image. First, a sample dataset containing text and image pairs is collected. These data pairs should describe the same content or scene. A pre-trained text model is used to extract feature vectors from the text, and a pre-trained image model is used to extract feature vectors from the image. For each pair of feature vectors of text-image data, it is marked as a positive sample pair (similar), and other texts or images are randomly selected to pair with the current text or image to form a negative sample pair (dissimilar). The contrastive loss function is used to encourage the feature vectors of positive sample pairs to be closer in space and the feature vectors of negative sample pairs to be farther apart. Driven by the contrastive loss, the model is trained to optimize the feature representation so that similar text-image pairs are closer in the feature space. Through training, the model learns feature representations that can map text and images to the same semantic space, thereby achieving cross-modal alignment.

[0052] On the other hand, for multi-source data across institutions, the embodiments of the present application can utilize the Multi-Entity Alignment Module, contrastive learning and synonymous semantic clustering technology to achieve unified alignment of coding standards, terminology expressions, and behavioral granularity among different financial institutions. The specific steps are as follows.

[0053] In some embodiments of the present application, the method also includes: determining the data source of the multimodal data; when the data source represents that the text data in the multimodal data comes from different financial institutions, extracting target keywords from the text data of different financial institutions, wherein the target keywords include at least one of the following: terminology expressions, specific codes, and behavior descriptions within each financial institution; according to the standard expression mapping table, converting the keywords corresponding to different financial institutions into a unified standard vocabulary, wherein the standard expression mapping table is a mapping table containing special terms of different financial institutions to a unified standard vocabulary, and the special terms of different financial institutions include the data expressions, feature codes, and behavior descriptions of each financial institution.

[0054] Specifically, raw data is collected from various financial institutions, including coding standards, terminology, and behavioral descriptions. This data is then cleaned to remove invalid or erroneous data to ensure data quality. The cleaned data is then annotated, including term standardization and behavioral classification, to facilitate subsequent processing. A term mapping table (standard expression mapping table) is then established based on this standardized data, mapping the terminology of different institutions to a unified terminology system. During this process, polysemous or synonymous terms can be disambiguated through contextual analysis to ensure terminology accuracy.

[0055] After acquiring new multi-source data, natural language processing techniques or deep learning-based keyword extraction algorithms can be used to automatically identify and extract keywords from text data. The extracted keywords are then contextually analyzed to ensure their accuracy and relevance. A term mapping table (standard expression mapping table) can be used to convert the extracted keywords into a unified standard vocabulary. Furthermore, as financial institutions update their internal terminology and financial industry terminology evolves, embodiments of the present application can continuously update the mapping table to maintain its timeliness and accuracy.

[0056] By building a standard expression mapping table, the unique expressions of various financial institutions are uniformly converted, eliminating the barriers to understanding caused by language differences and ensuring the consistency and reliability of knowledge graph construction. This approach not only improves data processing efficiency but also enhances cross-institutional collaboration, laying the foundation for building a more comprehensive and accurate transaction monitoring system.

[0057] The embodiments of this application can effectively semantically align digital RMB transaction data from different sources and modalities, constructing a dynamic knowledge graph that comprehensively reflects transaction activities. This process not only covers structured data, but also deeply analyzes unstructured data, thereby achieving a comprehensive understanding and monitoring of digital RMB transaction events. By mapping this data into a unified semantic space, the semantic gap between heterogeneous data is eliminated, allowing data from different modalities to be compared and analyzed within the same framework, significantly improving the accuracy and efficiency of abnormal behavior detection.

[0058] After obtaining the semantically aligned target data, we can continuously extract target triples from the target data and automatically construct an entity relationship network (for example, automatically generate multi-dimensional entity relationship networks such as user-account-device-behavior-contract), providing dynamic input for financial knowledge modeling driven by the target knowledge graph. The specific steps are as follows.

[0059] In some embodiments of the present application, constructing an entity relationship network based on target triples to obtain a target knowledge graph includes: using a stream processing engine and message middleware to monitor the transaction event stream, and when the multimodal data corresponding to the transaction event stream is updated, extracting the latest target triples from the target data corresponding to the updated multimodal data; based on the latest target triples, performing entity change perception and relationship evolution detection to obtain change information corresponding to the constructed target knowledge graph, wherein entity change perception and relationship evolution detection are used to determine the changed entity elements and / or relationship elements in the latest target triples compared to the constructed target knowledge graph; based on the change information, performing an update operation on the constructed target knowledge graph to obtain the latest target knowledge graph, wherein the update operation includes: node insertion, edge creation, and attribute update.

[0060] This graph construction mechanism does not rely on a static rule base, but supports continuous learning based on semantic space. It can automatically identify new concepts, fill in new relationships and update node weights as the scenario evolves, thereby forming a highly dynamic and highly evolvable knowledge base, providing structural guarantees for upper-level tasks such as smart contract tracing, risk control modeling, and compliance auditing.

[0061] Specifically, the present application embodiment designs a knowledge graph construction and evolution mechanism triggered by event streams, enabling real-time mapping of various changes in the digital RMB ecosystem, including transaction behavior, account structure, device changes, and contract execution. The system combines graph databases (such as Neo4j) with stream processing platforms such as Flink to build a second-level graph update channel, supporting version backtracking, subgraph differential annotation, and graph expansion. Simultaneously, this module integrates with the policy structure parsed by the large model to transform policy semantics into graph constraint nodes, thereby constructing a reasonable, multi-integrated knowledge graph system.

[0062] For example, in the knowledge graph construction and evolution mechanism triggered by event streams, we can first integrate stream processing engines (such as Apache Flink) and message middleware (such as Kafka) to achieve real-time capture and analysis of multi-source event streams from trading systems, account systems, and contract execution platforms, and perform preliminary filtering and classification based on event types, timestamps, and entity identifiers; in terms of incremental graph updates, a dual-track update strategy based on entity change perception and relationship evolution detection is designed. When new entities, new relationships, or changes in existing entity attributes are identified, the system automatically completes node insertion, edge establishment, or attribute updates in the graph database to ensure real-time synchronization of the graph structure and financial behavior status.

[0063] At the same time, a multi-granularity version control mechanism can also be introduced to record graph snapshots in time windows such as minutes, hours, and days, and cooperate with the dynamic evolution algorithm of behavioral labels based on node activity and evolution frequency to continuously optimize the semantic expression and risk labeling of entity nodes and relationship chains. For the policy structure obtained by large-scale model analysis, the system extracts the constraint entities, conditional rules, and applicable behaviors involved in the policy through the semantic parsing engine, and automatically generates corresponding policy nodes and constraint edges, which are injected into the graph to form a "policy-entity-behavior" ternary relationship structure. Further combined with subgraph expansion and policy influence propagation algorithms, the entity links and behavioral paths triggered by the policy are excavated, and a reasonable and traceable policy propagation chain is constructed to achieve dynamic perception and intelligent reasoning support for regulatory policies in financial behavior networks.

[0064] Real-time monitoring and rapid response are achieved through the stream processing engine and message middleware, ensuring the dynamic update capability of the knowledge graph. In the context of digital RMB transaction monitoring, this means that the system can promptly capture changes in any transaction event, whether it is the opening of a new account, changes in transaction amounts, or updates to contract status, all of which can be quickly reflected in the knowledge graph, providing a solid foundation for real-time detection of abnormal behavior. The entity change perception and relationship evolution detection mechanism can automatically identify the need for knowledge graph updates, reducing manual intervention and improving the system's automation level and response speed.

[0065] Furthermore, in the embodiments of the present application, graph neural networks and large model technologies can be integrated to establish a graph-enhanced semantic reasoning framework, fully utilize graph structured data to enhance knowledge relationships and entity features, and combine risk management rules to perform intelligent risk identification on the target knowledge graph. The specific steps are as follows.

[0066] In some embodiments of the present application, a graph-enhanced reasoning model is used in combination with risk management rules to identify abnormal paths in a target knowledge graph, including the following steps: in the graph neural network of the graph-enhanced reasoning model, a neighbor aggregation mechanism and a relationship perception mechanism are used to analyze the feature information of entity nodes and edges between entity nodes in the target knowledge graph, and generate node representations corresponding to each entity node containing contextual information, wherein the neighbor aggregation mechanism is used to integrate the feature information of the entity node and the neighbor nodes corresponding to the entity node, and the relationship perception mechanism is used to model the impact of different types of edges on information transmission, the feature information of the entity node includes at least one of the following: account attributes, transaction behavior characteristics, and contract status information, and the feature information of the edge includes at least one of the following: transaction amount, time interval, and call relationship; based on the node representation corresponding to each entity node in the target knowledge graph, the behavior pattern corresponding to each node path in the target knowledge graph is identified, wherein the behavior pattern includes at least one of the following: account fund jump, multi-contract call chain, and multi-account association control; using the large language model in the graph-enhanced reasoning model, according to the risk management rules corresponding to each behavior pattern, logical constraint verification and conflict detection are performed on the node path to obtain the risk probability corresponding to each node path, and the node path whose risk probability exceeds the preset risk probability is determined as an abnormal path.

[0067] In this embodiment, the graph-enhanced reasoning model can be used to identify complex and high-risk behaviors including cross-chain fund transfers, multi-account association control, and abnormal paths of smart contracts, such as Figure 3As shown in the figure, a model reasoning logic supporting the three-step identification chain of "path clustering-feature attribution-risk confidence" is constructed, which is suitable for multi-dimensional, nonlinear, and dynamically evolving financial risk control scenarios. Specifically, through entity feature encoding and relationship structure modeling, financial entity nodes such as accounts, transactions, and contracts and their associated relationships can be mapped into a heterogeneous graph. Node features include account attributes, transaction behavior characteristics, contract status information, etc., while edge features include multi-dimensional attributes such as transaction amount, time interval, and call relationship. During the graph neural network processing process, neighbor aggregation and relationship-aware aggregation mechanisms can be used to combine the node's own characteristics with the semantic information of adjacent nodes and edges. Node representations can be dynamically updated through multi-hop context propagation, effectively capturing complex behavior patterns such as cross-account fund jumps and multi-contract call chains, thereby enhancing the perception and reasoning capabilities of potential risky behaviors. Afterwards, graph search algorithms and large language models are used, combined with risk management rules, to model potential abnormal paths, perform logical constraint verification and conflict detection based on the graph rule engine, and make rule triggering judgments based on path and context conditions to achieve dynamic evaluation of risk control strategies, in order to identify risky behaviors including "splitting contract funds to evade supervision", "cross-contract jumps to hide the controlling party", and "function reentrancy vulnerabilities triggering fund migration".

[0068] In addition, the embodiments of the present application can realize the automatic update of risk control rules, and can continuously iterate and update risk identification features and patterns based on real-time data flow, user feedback, and changes in expert rules or policies, effectively adapting to the rapid evolution of new risk behaviors, and ensuring the system's continuous and efficient risk identification capabilities. The specific steps are as follows.

[0069] In some embodiments of the present application, the method also includes the following steps: when an update of the policy text related to the digital RMB is detected in the information source, obtaining the latest policy text; using a large language model to analyze the policy text to identify the constraint entities in the policy text, as well as the restrictions and business logic corresponding to the constraint entities; based on the restrictions and business logic, generating a grammatical constraint graph structure corresponding to the constraint entities, and converting the grammatical constraint graph structure into corresponding structured risk management rules.

[0070] For example, policy documents issued by financial regulators can be monitored in real time. Once policy updates are detected, multi-layer semantic understanding can be performed on newly released policy documents, automatically identifying the constrained entities (e.g., "corporate wallets" and "overseas payment accounts"), limiting conditions (e.g., "daily limits" and "cross-border operations require permission"), and business logic (e.g., "contract functions cannot be activated before real-name authentication"). A grammatical graph is generated based on a tree-structured parsing model, and then a graph-to-rule engine is used to express the policy structure. The parsed policy rules can be directly converted into control structures such as risk control rules and graph constraints within the system, and automatically injected into the system by business modules (e.g., payment system, contract engine, account system).

[0071] By automatically parsing policy texts using large language models, changes in policies and regulations can be captured in a timely manner, and updated risk management rules can be automatically generated. This is of great significance to the compliance operation of the digital RMB system. This method not only reduces the workload of manual interpretation of policy texts, but also ensures the real-time and accuracy of system rules, enabling rapid adaptation to new regulatory environments and avoiding security vulnerabilities caused by lagging rules. By converting the grammatical constraint graph structure into structured risk management rules, the system can more flexibly adjust its monitoring strategy and improve the recognition rate of abnormal transactions.

[0072] At the same time, in order to improve the interpretability of risk identification results, the system can also introduce the SHAP (SHapley Additive exPlanations) mechanism after the graph neural network reasoning is completed. By using the graph visualization engine and large model language generation capabilities, based on the game theory feature attribution method, it quantifies the influence of each node feature, neighbor contribution and path structure on the final risk judgment, and automatically outputs the risk path, graph structure evidence chain and reasoning explanation text to support the output of risk propagation chain visualization map and feature attribution report, ensuring that the risk decision-making process is auditable, traceable and interpretable, thereby improving the interpretability of the risk control decision-making process and solving the problem of lack of auditability in related technologies.

[0073] For example, the content of the structured risk control report automatically generated by the system may include but is not limited to: 1) risk assessment level (high / medium / low risk and its confidence level); 2) trigger path map (visual graph of contract node call chain and abnormal trigger path); 3) risk factor identification description (trigger function, variable value, call context, etc.); 4) policy hit clauses (entries corresponding to regulatory policies); 5) generative natural language explanation (for auditors to read).

[0074] It should be noted that, given the multi-institutional nature of the digital RMB (e.g., central banks, commercial banks, payment institutions, and clearing platforms), the platform or system implementing the methods described in the embodiments of this application incorporates a tenant isolation mechanism. This mechanism utilizes namespace partitioning, policy-based access control (RBAC), API throttling, and resource quota policies to ensure data security, policy independence, and service isolation among various business entities. Furthermore, differentiated deployment models (centralized, edge, and hybrid cloud deployment) are supported to accommodate the regulatory and compliance requirements of various financial institutions.

[0075] The digital RMB-related data processing method based on large model technology proposed in this application realizes the effective fusion and semantic alignment of multimodal data through a cross-modal embedding model and a unified semantic space construction mechanism, constructs a dynamically updated knowledge graph, and then uses a graph-enhanced reasoning model combined with risk management rules to monitor and identify abnormal behaviors in digital RMB transactions in real time. This method not only improves the accuracy and efficiency of transaction monitoring, but also can adapt to the ever-changing financial environment and policy requirements, and enhance the security and compliance of the digital RMB system. In addition, by standardizing the special terms of different financial institutions, it promotes information sharing and understanding across institutions, further improving the collaborative efficiency of the entire financial ecosystem.

[0076] According to an embodiment of the present application, an embodiment of a device for processing digital RMB related data based on big model technology is also provided. Figure 4 This is a structural diagram of a device for processing digital RMB related data based on large model technology according to an embodiment of the present application. Figure 4 As shown, the device includes:

[0077] The semantic alignment module 40 is used to obtain the multimodal data involved in the transaction event stream of the digital RMB and map the multimodal data into a unified semantic space for semantic alignment to obtain target data. In the unified semantic space, the vectors corresponding to the more similar the semantics of the data are, the closer the distance is, and the vectors corresponding to the data with greater semantic differences are, the farther the distance is.

[0078] A network construction module 42 is configured to extract target triples from the target data and construct an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to represent the association relationship between two entities;

[0079] The risk identification module 44 is used to use a graph-enhanced reasoning model, combined with risk management rules, to identify abnormal paths in the target knowledge graph, where the abnormal paths correspond to abnormal processing behaviors of the digital RMB. The risk management rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behaviors of the digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and large model technologies.

[0080] Optionally, the multimodal data includes: structured data and unstructured data of different modalities, wherein the structured data includes at least one of the following: transaction records, account information, device binding records, and the unstructured data includes at least one of the following: policy text, user behavior logs, contract descriptions, and image credentials; mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data includes: using a modal encoder in a semantic alignment model to extract semantic features corresponding to data of different modalities, wherein the feature extraction methods corresponding to data of different modalities are different, and the modalities include at least one of the following: text, voice, image, and video; using a connector in a semantic alignment model to convert the semantic features corresponding to different modalities, wherein the conversion processing is used to eliminate the differences in distribution, scale, and noise of semantic features of different modalities, and obtain semantic features in a unified representation form to ensure that data of all modalities can be represented in the same semantic space; using a generator in a semantic alignment model to map the converted semantic features to a unified semantic space to obtain target data.

[0081] Optionally, the training steps of the semantic alignment model include: obtaining a sample data set, wherein the sample data set contains multiple sample data, and each sample data contains data of different modalities for describing the same content; performing feature extraction on the data of different modalities in the sample data to obtain feature vectors corresponding to different modalities; pairing the feature vectors of different modalities corresponding to the same sample data to obtain positive sample pairs, and randomly selecting feature vectors of different modalities corresponding to different sample data to pair to obtain negative sample pairs; forming a training data set with positive samples and negative sample pairs, and based on the training data set, training the initial model in combination with the contrast loss function to obtain a semantic alignment model, wherein the contrast loss function is used to make the feature vectors in the positive sample pairs in the results output by the initial model closer in the semantic space, while the feature vectors in the negative sample pairs are farther apart in the semantic space during the training process.

[0082] Optionally, the semantic alignment module 40 is also used to: determine the data source of the multimodal data; when the data source represents that the text data in the multimodal data comes from different financial institutions, extract target keywords from the text data of different financial institutions, wherein the target keywords include at least one of the following: terminology expressions, specific codes, and behavior descriptions within each financial institution; according to the standard expression mapping table, convert the keywords corresponding to different financial institutions into a unified standard vocabulary, wherein the standard expression mapping table is a mapping table containing special terms of different financial institutions to a unified standard vocabulary, and the special terms of different financial institutions include the data expressions, feature codes, and behavior descriptions of each financial institution.

[0083] Optionally, constructing an entity relationship network based on the target triples to obtain a target knowledge graph includes: using a stream processing engine and message middleware to monitor the transaction event stream, and when the multimodal data corresponding to the transaction event stream is updated, extracting the latest target triples from the target data corresponding to the updated multimodal data; based on the latest target triples, performing entity change perception and relationship evolution detection to obtain change information corresponding to the constructed target knowledge graph, wherein entity change perception and relationship evolution detection are used to determine the changed entity elements and / or relationship elements in the latest target triples compared to the constructed target knowledge graph; based on the change information, performing an update operation on the constructed target knowledge graph to obtain the latest target knowledge graph, wherein the update operation includes: node insertion, edge creation, and attribute update.

[0084] Optionally, using a graph-enhanced reasoning model in combination with risk management rules to identify abnormal paths in a target knowledge graph includes: in a graph neural network of the graph-enhanced reasoning model, using a neighbor aggregation mechanism and a relationship perception mechanism to analyze feature information of entity nodes and edges between entity nodes in the target knowledge graph, and generating a node representation corresponding to each entity node containing contextual information, wherein the neighbor aggregation mechanism is used to integrate feature information of entity nodes and neighbor nodes corresponding to the entity nodes, and the relationship perception mechanism is used to model the impact of different types of edges on information transmission, the feature information of the entity nodes includes at least one of the following: account attributes, transaction behavior characteristics, and contract status information, and the feature information of the edges includes at least one of the following: transaction amount, time interval, and call relationship; based on the node representation corresponding to each entity node in the target knowledge graph, identifying the behavior pattern corresponding to each node path in the target knowledge graph, wherein the behavior pattern includes at least one of the following: account fund jump, multi-contract call chain, and multi-account association control; using a large language model in the graph-enhanced reasoning model, based on the risk management rules corresponding to each behavior pattern, performing logical constraint verification and conflict detection on the node path, obtaining the risk probability corresponding to each node path, and determining the node path whose risk probability exceeds the preset risk probability as an abnormal path.

[0085] Optionally, the risk identification module 44 is also used to: obtain the latest policy text when an update of the policy text related to the digital RMB is detected in the information source; use a large language model to analyze the policy text to identify the constraint entities in the policy text, as well as the restrictions and business logic corresponding to the constraint entities; generate a grammatical constraint graph structure corresponding to the constraint entities based on the restrictions and business logic, and convert the grammatical constraint graph structure into corresponding structured risk management rules.

[0086] It should be noted that the various modules in the above-mentioned digital RMB-related data processing device based on big model technology can be a program module (for example, a set of program instructions to implement a certain specific function) or a hardware module. For the latter, it can be expressed in the following form, but not limited to this: the expression form of each of the above-mentioned modules is a processor, or the functions of each of the above-mentioned modules are implemented by a processor.

[0087] It should be noted that the processing device for digital RMB related data based on the large model technology provided in this embodiment can be used to execute Figure 2 The method for processing digital RMB related data based on the big model technology shown, therefore, the relevant explanations of the above-mentioned method for processing digital RMB related data based on the big model technology are also applicable to the embodiments of this application and will not be repeated here.

[0088] An embodiment of the present application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the following method for processing digital RMB-related data based on big model technology by running the computer program: obtaining multimodal data involved in the transaction event stream of digital RMB, and mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data, wherein in the unified semantic space, the more semantically similar the data, the closer the distance between the vectors corresponding to the data, and the greater the semantic difference, the farther the distance between the vectors corresponding to the data; extracting target triples from the target data, and constructing an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to characterize the association relationship between two entities; using a graph-enhanced reasoning model, combined with risk control rules, to identify abnormal paths in the target knowledge graph, wherein the abnormal paths correspond to abnormal processing behaviors of digital RMB, and the risk control rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behaviors of digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and big model technology.

[0089] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method for processing digital RMB-related data based on big model technology described in each embodiment of the present application: obtaining multimodal data involved in the transaction event stream of digital RMB, and mapping the multimodal data to a unified semantic space for semantic alignment to obtain target data, wherein, in the unified semantic space, the vectors corresponding to data with more similar semantics are closer, and the vectors corresponding to data with greater semantic differences are farther; extracting target triples from the target data, and constructing an entity relationship network based on the target triples to obtain a target knowledge graph, wherein the target triples contain two entity elements and one relationship element, and the target triples are used to characterize the association relationship between two entities; using a graph-enhanced reasoning model, combined with risk management rules, to identify abnormal paths in the target knowledge graph, wherein the abnormal paths correspond to abnormal processing behaviors of digital RMB, and the risk management rules are used to indicate the characteristics and patterns of rules that should be followed for the processing behaviors of digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and big model technology.

[0090] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0091] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0092] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0093] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0094] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0095] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program code.

[0096] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for processing digital RMB related data based on large model technology, characterized in that: include: Obtain multimodal data involved in the transaction event stream of the digital RMB, and map the multimodal data into a unified semantic space for semantic alignment to obtain target data, wherein, in the unified semantic space, the vectors corresponding to data with more similar semantics are closer in distance, and the vectors corresponding to data with greater semantic differences are farther in distance; Extracting a target triple from the target data, and constructing an entity relationship network based on the target triple to obtain a target knowledge graph, wherein the target triple contains two entity elements and one relationship element, and the target triple is used to represent the association relationship between two entities; A graph-enhanced reasoning model is used in combination with risk management rules to identify abnormal paths in the target knowledge graph, including: in the graph neural network of the graph-enhanced reasoning model, a neighbor aggregation mechanism and a relationship perception mechanism are used to analyze the characteristic information of entity nodes and edges between entity nodes in the target knowledge graph, and generate node representations containing context information corresponding to each entity node, wherein the neighbor aggregation mechanism is used to integrate the characteristic information of entity nodes and neighbor nodes corresponding to the entity nodes, and the relationship perception mechanism is used to model the impact of different types of edges on information transmission, the characteristic information of the entity nodes includes at least one of the following: account attributes, transaction behavior characteristics, contract status information, and the characteristic information of the edges includes at least one of the following: transaction amount, time interval, call relationship; based on each of the target knowledge graphs The node representation corresponding to the entity node identifies the behavior pattern corresponding to each node path in the target knowledge graph, wherein the behavior pattern includes at least one of the following: account funds jump, multi-contract call chain, and multi-account association control; using the large language model in the graph-enhanced reasoning model, according to the risk control rules corresponding to each of the behavior patterns, the node path is subjected to logical constraint verification and conflict detection to obtain the risk probability corresponding to each of the node paths, and the node path whose risk probability exceeds the preset risk probability is determined as the abnormal path; wherein, the abnormal path corresponds to the abnormal processing behavior of the digital RMB, and the risk control rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behavior of the digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and large model technologies.

2. The method for processing digital RMB related data based on large model technology according to claim 1 is characterized in that: The multimodal data includes: structured data and unstructured data of different modalities, wherein the structured data includes at least one of the following: transaction records, account information, and device binding records; the unstructured data includes at least one of the following: policy text, user behavior logs, contract descriptions, and image credentials; the multimodal data is mapped into a unified semantic space for semantic alignment, and the target data obtained includes: Using a modality encoder in a semantic alignment model to extract semantic features corresponding to data of different modalities, wherein the features corresponding to data of different modalities are extracted in different ways, and the modalities include at least one of the following: text, speech, image, and video; Using the connector in the semantic alignment model, the semantic features corresponding to different modalities are converted, wherein the conversion process is used to eliminate differences in distribution, scale, and noise between the semantic features of different modalities and obtain the semantic features in a unified representation form to ensure that data from all modalities can be represented in the same semantic space; The generator in the semantic alignment model is used to map the semantic features after the conversion process into the unified semantic space to obtain the target data.

3. The method for processing digital RMB related data based on large model technology according to claim 2 is characterized in that: The training steps of the semantic alignment model include: Acquire a sample data set, wherein the sample data set includes multiple pieces of sample data, and each piece of sample data includes data in different modes for describing the same content; Extracting features from data of different modes in the sample data to obtain feature vectors corresponding to the different modes; Pairing the feature vectors of different modes corresponding to the same piece of sample data to obtain a positive sample pair, and randomly selecting the feature vectors of different modes corresponding to different pieces of sample data to pair them to obtain a negative sample pair; The positive samples and the negative sample pairs are composed of a training data set, and based on the training data set, the initial model is trained in combination with a contrast loss function to obtain the semantic alignment model, wherein the contrast loss function is used to make the feature vectors of the positive sample pairs in the results output by the initial model closer in the semantic space, and the feature vectors of the negative sample pairs farther apart in the semantic space during the training process.

4. The method for processing digital RMB related data based on large model technology according to claim 2 is characterized in that: The method further comprises: Determining a data source of the multimodal data; In a case where the data source represents that the text data in the multimodal data originates from different financial institutions, extracting target keywords from the text data of different financial institutions, wherein the target keywords include at least one of the following: terminology, specific codes, and behavior descriptions within each financial institution; According to the standard expression mapping table, the keywords corresponding to the different financial institutions are converted into a unified standard vocabulary, wherein the standard expression mapping table is a mapping table containing special terms of different financial institutions to unified standard vocabulary, and the special terms of different financial institutions include the data expression, feature coding, and behavior description of each financial institution.

5. The method for processing digital RMB related data based on large model technology according to claim 1 is characterized in that: An entity relationship network is constructed based on the target triples to obtain a target knowledge graph including: Using a stream processing engine and message middleware, monitoring the transaction event stream, and when the multimodal data corresponding to the transaction event stream is updated, extracting the latest target triple from the target data corresponding to the updated multimodal data; Based on the latest target triple, entity change perception and relationship evolution detection are performed to obtain change information corresponding to the constructed target knowledge graph, wherein the entity change perception and relationship evolution detection are used to determine the entity elements and / or relationship elements that have changed in the latest target triple compared to the constructed target knowledge graph; Based on the change information, the target knowledge graph that has been constructed is updated to obtain the latest target knowledge graph, wherein the update operation includes: node insertion, edge creation, and attribute update.

6. The method for processing digital RMB related data based on large model technology according to claim 1 is characterized in that: The method further comprises: Upon detecting an update of a policy text related to the digital RMB in an information source, obtain the latest policy text; Using a large language model, the policy text is analyzed to identify constraint entities in the policy text, as well as restrictions and business logic corresponding to the constraint entities; Based on the restriction conditions and the business logic, a grammatical constraint graph structure corresponding to the constraint entity is generated, and the grammatical constraint graph structure is converted into the corresponding structured risk management and control rules.

7. A device for processing digital RMB related data based on large model technology, characterized in that: include: A semantic alignment module is used to obtain multimodal data involved in the transaction event stream of the digital RMB, and map the multimodal data into a unified semantic space for semantic alignment to obtain target data. In the unified semantic space, the vectors corresponding to data with more similar semantics are closer, and the vectors corresponding to data with greater semantic differences are farther apart. A network construction module is used to extract a target triple from the target data and construct an entity relationship network based on the target triple to obtain a target knowledge graph, wherein the target triple contains two entity elements and one relationship element, and the target triple is used to represent the association relationship between two entities; A risk identification module is used to use a graph-enhanced reasoning model in combination with risk management rules to identify abnormal paths in the target knowledge graph, including: in the graph neural network of the graph-enhanced reasoning model, a neighbor aggregation mechanism and a relationship perception mechanism are used to analyze the characteristic information of entity nodes and edges between entity nodes in the target knowledge graph, and generate node representations corresponding to each entity node containing context information, wherein the neighbor aggregation mechanism is used to integrate the characteristic information of entity nodes and neighbor nodes corresponding to the entity nodes, and the relationship perception mechanism is used to model the impact of different types of edges on information transmission, the characteristic information of the entity nodes includes at least one of the following: account attributes, transaction behavior characteristics, contract status information, and the characteristic information of the edges includes at least one of the following: transaction amount, time interval, call relationship; based on the target knowledge graph The node representation corresponding to each of the entity nodes in the target knowledge graph identifies the behavior pattern corresponding to each node path in the target knowledge graph, wherein the behavior pattern includes at least one of the following: account funds jump, multi-contract call chain, and multi-account association control; using the large language model in the graph-enhanced reasoning model, according to the risk control rules corresponding to each of the behavior patterns, the node path is subjected to logical constraint verification and conflict detection to obtain the risk probability corresponding to each of the node paths, and the node path whose risk probability exceeds the preset risk probability is determined as the abnormal path; wherein, the abnormal path corresponds to the abnormal processing behavior of the digital RMB, and the risk control rules are used to indicate the characteristics and patterns of the rules that should be followed for the processing behavior of the digital RMB. The graph-enhanced reasoning model is an inference framework that combines graph neural networks and large model technologies.

8. An electronic device, characterized in that: include: A memory and a processor, the processor being used to run a program stored in the memory, wherein the program, when running, executes the method for processing digital RMB related data based on large model technology as described in any one of claims 1 to 6.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the method for processing digital RMB related data based on large model technology as described in any one of claims 1 to 6 by running the computer program.

Images