Credit clause auditing method and device, computer equipment and storage medium

Abstract

The invention discloses a credit clause auditing method and device, computer equipment and a storage medium, and belongs to the technical field of artificial intelligence. Firstly, a credit clause text is analyzed through a natural language processing technology; performing association analysis on the named entities by using the domain dictionary, and determining the dependency relationship between the entities; then, key information features in the named entities and the dependency relationship thereof are deeply extracted and learned; and constructing a compliance judgment knowledge graph, and mapping the extracted key information features to corresponding entities and relationships in the graph to form a structured credit clause representation. And finally, evaluating the compliance of the credit clauses by combining a compliance judgment knowledge graph through a constructed auditing rule engine. The invention also relates to the technical field of block chains, and the credit clause text is stored on a block chain node. According to the method and the device, whether the credit clauses meet the preset compliance requirements or not is automatically evaluated, so that efficient and accurate auditing of the credit clauses is realized.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and specifically relates to a method, device, computer equipment and storage medium for reviewing letter of credit terms. Background Art

[0002] The review of letter of credit terms is a key link in the bank's trade financing business. It requires the rapid and accurate extraction of key information from a large amount of letter of credit terms texts and the review in accordance with international practices and internal bank rules.

[0003] The language of letter of credit terms is diverse and the content is complicated. The formats and terms of letters of credit vary from bank to bank and from country to country, which brings great challenges to information extraction. In addition, letter of credit terms often contain a large number of business terms, abbreviations, typos, etc., which need to be standardized during the information extraction process. The review of letter of credit terms also needs to combine international practices and internal bank rules to make compliance judgments on the extracted key information, which places high demands on the knowledge integration capabilities of the algorithm model.

[0004] At present, the review of letter of credit terms mainly relies on manual operation, which has problems such as low efficiency, high error rate, and inconsistent review standards. Therefore, how to use artificial intelligence technology, especially the information extraction ability of large models, to realize the automated review of letter of credit terms is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a method, device, computer equipment and storage medium for reviewing letter of credit terms, so as to solve the problems that the existing letter of credit terms review mainly relies on manual operation, has low efficiency, high error rate, inconsistent review standards, etc.

[0006] In order to solve the above technical problems, the present application embodiment provides a method for reviewing letter of credit terms, which adopts the following technical solution:

[0007] A method for reviewing letter of credit terms, comprising:

[0008] Obtain the text of the letter of credit terms and perform natural language processing on the text of the letter of credit terms, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition;

[0009] According to the pre-established domain dictionary, the named entities in the letter of credit clause text are analyzed for association and the dependency relationship of the named entities is obtained;

[0010] A neural network model is used to extract and learn features of named entities and dependencies to learn key information features in letter of credit terms.

[0011] Construct a compliance judgment knowledge graph for letter of credit terms and map key information features to entities and relationships in the compliance judgment knowledge graph;

[0012] Build an audit rule engine for letters of credit, and determine whether the terms of the letter of credit meet compliance requirements based on the audit rule engine and the compliance judgment knowledge graph.

[0013] Furthermore, the steps of obtaining the text of the letter of credit terms and performing natural language processing on the text of the letter of credit terms specifically include:

[0014] Obtain the text of the letter of credit terms, and perform data structure processing on the text of the letter of credit terms to obtain structured text data;

[0015] Perform word segmentation operation on structured text data to obtain text segmentation;

[0016] Perform part-of-speech tagging on text segmentation words and identify the part-of-speech information of text segmentation words;

[0017] According to the part-of-speech information of text segmentation, named entity recognition technology is used to identify and extract named entities in the text of letter of credit terms.

[0018] Furthermore, according to the pre-established domain dictionary, the named entities in the text of the letter of credit terms are subjected to association analysis to obtain the dependency relationship of the named entities, which specifically includes:

[0019] Determine the business domain to which the named entity belongs based on the domain category defined in the domain dictionary;

[0020] Identify the entity start position and entity end position of the named entity, and obtain the position information of the named entity in the text of the letter of credit terms;

[0021] Based on the business domain to which the named entities belong and the location information of the named entities in the text of the letter of credit terms, the named entities are associated and analyzed to obtain the dependency relationships of the named entities.

[0022] Further, the step of performing association analysis on the named entities based on the business fields to which the named entities belong and the position information of the named entities in the text of the letter of credit terms to obtain the dependency relationship of the named entities specifically includes:

[0023] Determine the associated entity that has semantic association with the named entity according to the business field to which the named entity belongs and the position information of the named entity in the text of the letter of credit clause, wherein the semantic association includes subject-predicate association relationship, verb-object association relationship, and attributive-predicate association relationship;

[0024] Identify the entity start position and entity end position of the associated entity, and obtain the position information of the associated entity in the text of the letter of credit terms;

[0025] According to the position information of the named entity in the text of the letter of credit terms and the position information of the associated entity in the text of the letter of credit terms, a dependency relationship mapping diagram between entities is established;

[0026] Based on the dependency mapping graph between entities, the dependency relationships of named entities are generated.

[0027] Furthermore, a neural network model is used to extract and learn features of named entities and dependencies to learn key information features in the terms of the letter of credit, including:

[0028] Combine the named entities and the dependency relationships to obtain a named entity-dependency relationship combination;

[0029] Use the preset word vector model to vectorize the named entity-dependency combination to obtain the combined word vector;

[0030] Use the neural network model to learn the features of the combined word vectors, and iteratively update the neural network model based on the preset optimization algorithm until the model is fitted;

[0031] After the model is fitted, the feature learning results output by the neural network model are obtained to obtain key information features.

[0032] Furthermore, the key information features include several groups of named entity-dependency combinations, and the steps of constructing a compliance judgment knowledge graph of the letter of credit terms and mapping the key information features to entities and relationships in the compliance judgment knowledge graph specifically include:

[0033] Cluster the business domains to which the named entities belong and obtain domain cluster labels;

[0034] Determine the letter of credit review template that matches the letter of credit clause text based on the domain clustering labels;

[0035] Constructing a compliance judgment knowledge graph of the letter of credit terms based on the letter of credit review template, wherein the compliance judgment knowledge graph includes nodes and edges;

[0036] The named entities are matched to the nodes of the compliance judgment knowledge graph according to the ontology matching algorithm, and the dependency relationships are mapped to the edges of the compliance judgment knowledge graph according to the relationship mapping rules.

[0037] Furthermore, a review rule engine for the letter of credit is constructed, and the steps of judging whether the terms of the letter of credit meet the compliance requirements according to the review rule engine and the compliance judgment knowledge graph specifically include:

[0038] Obtain the review rules in the letter of credit review template, and use the decision tree algorithm to establish a rule model to obtain the review rule engine;

[0039] Extract key features of the letter of credit clause text to obtain key features of the text;

[0040] Use the audit rule engine to conduct compliance audit on key features of the text based on the random forest algorithm to obtain the initial non-compliant items in the text of the letter of credit terms;

[0041] Using knowledge graph reasoning technology, multi-hop reasoning is performed on the initial non-compliant items in the compliance judgment knowledge graph to obtain multi-hop reasoning results;

[0042] Based on the multi-hop reasoning results, the probability that the initial non-compliant items of the letter of credit meet the compliance requirements is calculated, and the non-compliant items are screened using the preset probability threshold to obtain the non-compliant items of the letter of credit terms.

[0043] In order to solve the above technical problems, the present application also provides a device for reviewing letter of credit terms, which adopts the following technical solution:

[0044] A device for reviewing letter of credit terms, comprising:

[0045] A text processing module is used to obtain the text of the letter of credit terms and perform natural language processing on the text of the letter of credit terms, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition;

[0046] The association analysis module is used to perform association analysis on the named entities in the text of the letter of credit terms according to the pre-established domain dictionary to obtain the dependency relationship of the named entities;

[0047] A feature learning module is used to extract and learn features of named entities and dependency relationships using a neural network model to learn key information features in the terms of the letter of credit;

[0048] A graph construction module is used to construct a compliance judgment knowledge graph of letter of credit terms and map key information features to entities and relationships in the compliance judgment knowledge graph;

[0049] The terms review module is used to build a review rule engine for letters of credit, and to determine whether the terms of the letter of credit meet the compliance requirements based on the review rule engine and the compliance judgment knowledge graph.

[0050] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0051] A computer device comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the method for reviewing letter of credit terms as described in any one of the above are implemented.

[0052] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0053] A computer-readable storage medium having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by a processor, implement the steps of the method for reviewing letter of credit terms as described in any one of the above.

[0054] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0055] The present application discloses a method, device, computer equipment and storage medium for reviewing letter of credit terms, which belongs to the field of artificial intelligence technology. First, the text of the letter of credit terms is parsed by natural language processing technology. Then, the domain dictionary is used to perform association analysis of named entities to clarify the dependencies between entities. Then, the key information features in these named entities and their dependencies are deeply extracted and learned. A compliance judgment knowledge graph is constructed, and the extracted key information features are mapped to the corresponding entities and relationships in the graph to form a structured representation of the letter of credit terms. Finally, the constructed review rule engine is combined with the compliance judgment knowledge graph to evaluate the compliance of the letter of credit terms. Through automated processing and intelligent analysis, this application significantly improves the efficiency and accuracy of letter of credit review, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 An exemplary system architecture diagram is shown in which the present application can be applied;

[0058] Figure 2 A flow chart showing an embodiment of a method for reviewing letter of credit terms according to the present application is shown;

[0059] Figure 3 A schematic structural diagram of an embodiment of a device for reviewing letter of credit terms according to the present application is shown;

[0060] Figure 4 A schematic structural diagram of an embodiment of a computer device according to the present application is shown. DETAILED DESCRIPTION

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0062] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0063] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0064] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0065] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0066] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer and a desktop computer, etc.

[0067] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0068] It should be noted that the method for reviewing the terms of a letter of credit provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the device for reviewing the terms of a letter of credit is generally arranged in the server / terminal device.

[0069] It should be understood that Figure 1 The numbers of terminal devices, networks and servers in the embodiment are merely illustrative. The above system may have any number of terminal devices, networks and servers according to implementation requirements.

[0070] Continue to refer Figure 2 , showing a flow chart of an embodiment of the method for reviewing the terms of a letter of credit according to the present application. The present application embodiment can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0071] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning. The method for reviewing the terms of a letter of credit includes the following steps:

[0072] S201, obtaining the text of the terms of the letter of credit, and performing natural language processing on the text of the terms of the letter of credit, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition.

[0073] Letter of Credit (L / C) is a written guarantee document issued by a bank (issuing bank) to a beneficiary (usually a seller) based on the applicant's (usually the buyer's) request and instructions, with a certain amount of money to be paid within a certain period of time against documents that meet the requirements. It is one of the most important and commonly used payment methods in international trade, and it ensures the rights and interests of both parties to the transaction through bank credit.

[0074] The terms of a letter of credit usually include the number, type (such as sight, usance, etc.), applicant, beneficiary, redemption method and redemption bank, name of the goods, unit price, quantity, packaging, mode of transport, place of shipment, destination, latest shipment time, whether partial shipment and transshipment are allowed, etc.

[0075] Specifically, after obtaining the text of the letter of credit terms, the system uses natural language processing technology to deeply analyze the text of the letter of credit terms. Natural language processing is not limited to basic word segmentation, but also covers part-of-speech tagging to accurately identify the grammatical role of each word in the sentence. The system also introduces named entity recognition technology to extract entities with specific meanings from complex texts, such as bank names, amounts, dates and other key information.

[0076] S202, performing association analysis on named entities in the text of the letter of credit terms according to a pre-established domain dictionary to obtain dependency relationships of the named entities.

[0077] Specifically, the domain dictionary not only contains common entity vocabulary in the letter of credit business, but also defines the domain to which the entity belongs and the potential relationship between these entities. For example, through this dictionary, the system can intelligently identify key named entities in the terms, such as bank name, transaction amount, cargo description, etc., and further analyze the logic and dependency relationships between these entities. For example, the system can identify the execution relationship between a bank name and a payment instruction, or the matching relationship between a cargo description and a shipping clause. This deep association analysis enhances the depth of text understanding.

[0078] S203, using a neural network model to perform feature extraction and feature learning on named entities and dependency relationships to learn key information features in the terms of the letter of credit.

[0079] Specifically, in order to more accurately capture the key information features in the terms of the letter of credit, the system introduces an advanced neural network model for deep feature extraction and learning. The neural network model can automatically extract rich semantic features and contextual information from named entities and their dependencies, and perform nonlinear transformations and combinations through a multi-layer network structure to discover complex patterns and rules hidden in text data. In addition, the system also uses a variety of optimization algorithms to iteratively train the model to improve its feature learning capabilities and accuracy. In this way, the key information in the terms of the letter of credit can be understood more comprehensively and accurately.

[0080] S204, construct a compliance judgment knowledge graph for the letter of credit terms, and map key information features to entities and relationships in the compliance judgment knowledge graph.

[0081] Specifically, the system further combines technical depth with business understanding when constructing the compliance judgment knowledge graph of letter of credit terms. The knowledge graph is not only a network that stores entities and relationships, but also an intelligent framework that can reflect business logic and rules. By mapping key information features to the corresponding entities and relationships of the knowledge graph, the system can achieve accurate docking between letter of credit terms and compliance requirements. This process not only involves the technical challenges of entity recognition and relationship extraction, but also requires a deep understanding of the rules and processes of letter of credit business.

[0082] S205, build a review rule engine for the letter of credit, and determine whether the terms of the letter of credit meet the compliance requirements based on the review rule engine and the compliance judgment knowledge graph.

[0083] Specifically, in order to enhance the intelligence and accuracy of letter of credit review, the system needs to pre-build a rule engine for the review of letter of credit terms and deeply integrate it with the compliance judgment knowledge graph. This rule engine not only integrates a wealth of review rules and standards, but also has the ability to dynamically learn and adaptively adjust. During the letter of credit review process, the system first conducts a preliminary review based on the review rule engine. For non-compliant items obtained in the preliminary review, it continues to use the compliance judgment knowledge graph and combines multi-hop reasoning to obtain a more comprehensive compliance judgment result, and realizes a comprehensive and multi-dimensional compliance assessment of the letter of credit terms. This technical expansion not only improves the review efficiency, but also ensures the objectivity and accuracy of the review results.

[0084] In the above embodiments, the present application significantly improves the efficiency and accuracy of letter of credit review through automated processing and intelligent analysis, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results.

[0085] Furthermore, the steps of obtaining the text of the letter of credit terms and performing natural language processing on the text of the letter of credit terms specifically include:

[0086] Obtain the text of the letter of credit terms and conditions, and perform data structure processing on the text of the letter of credit terms and conditions to obtain structured text data;

[0087] Perform word segmentation operation on structured text data to obtain text segmentation;

[0088] Perform part-of-speech tagging on text segmentation words and identify the part-of-speech information of text segmentation words;

[0089] According to the part-of-speech information of text segmentation, named entity recognition technology is used to identify and extract named entities in the text of letter of credit terms.

[0090] In this embodiment, first, after the system obtains the original text data of the letter of credit terms, it needs to pre-process the text of the letter of credit terms. Regular expressions and other methods can be used to remove special characters, spaces, etc. in the text to obtain structured text data. Then, the pre-processed text data is input into the natural language processing module, and the text is segmented using a word segmentation tool to obtain a word segmentation result. Next, a part-of-speech tagging tool is used to tag the word segmentation result, and the part-of-speech information such as nouns, verbs, and adjectives in the text is identified. On this basis, a named entity recognition algorithm is used to identify and extract key named entities in the letter of credit text, such as the letter of credit number, the applicant for the letter of credit, and the beneficiary.

[0091] If the extracted entity information is incomplete, such as missing letter of credit number, applicant, beneficiary, etc., the missing fields will be returned to the system's preset data completion module, and the missing letter of credit information will be obtained by querying the database, crawling the web pages, etc., to ensure the integrity of the text information.

[0092] In the above embodiment, the automatic parsing and key information extraction of the letter of credit terms are realized by structural processing, word segmentation, part-of-speech tagging and named entity recognition of the letter of credit terms text, which effectively improves the accuracy and efficiency of letter of credit review and reduces the complexity and error rate of manual interpretation.

[0093] Furthermore, according to the pre-established domain dictionary, the named entities in the text of the letter of credit terms are subjected to association analysis to obtain the dependency relationship of the named entities, which specifically includes:

[0094] Determine the business domain to which the named entity belongs based on the domain category defined in the domain dictionary;

[0095] Identify the entity start position and entity end position of the named entity, and obtain the position information of the named entity in the text of the letter of credit terms;

[0096] Based on the business domain to which the named entities belong and the location information of the named entities in the text of the letter of credit terms, the named entities are associated and analyzed to obtain the dependency relationships of the named entities.

[0097] In this embodiment, the domain dictionary predefines multiple business domain categories, including finance, law, and medical care, to support accurate identification and association analysis of professional terms, and to improve the professionalism and accuracy of the review of letter of credit terms. First, the system matches the named entity through the domain dictionary, and determines the business domain to which the named entity belongs based on the predefined domain category label. For example, if the named entity "dollar" is identified, the system will search in the domain dictionary and classify it as the "finance" field. Then, the system obtains the position information of the named entity in the text of the letter of credit terms by identifying the entity start position and entity end position of the named entity in the text of the letter of credit terms. Finally, based on the business domain to which the named entity belongs and the position information of the named entity in the text of the letter of credit terms, the dependency relationship between the named entity and other entities is determined, and the possible semantic associations such as subject-predicate relationship, verb-object relationship, and attributive relationship are identified, thereby obtaining the dependency relationship of the named entity.

[0098] In the above embodiment, by combining the domain dictionary and text position information, a deep association analysis of the named entities in the terms of the letter of credit is achieved, and the dependency relationship between the named entities is accurately revealed.

[0099] Further, the step of performing association analysis on the named entities based on the business fields to which the named entities belong and the position information of the named entities in the text of the letter of credit terms to obtain the dependency relationship of the named entities specifically includes:

[0100] Determine the associated entity that has semantic association with the named entity according to the business field to which the named entity belongs and the position information of the named entity in the text of the letter of credit clause, wherein the semantic association includes subject-predicate association relationship, verb-object association relationship, and attributive-predicate association relationship;

[0101] Identify the entity start position and entity end position of the associated entity, and obtain the position information of the associated entity in the text of the letter of credit terms;

[0102] According to the position information of the named entity in the text of the letter of credit terms and the position information of the associated entity in the text of the letter of credit terms, a dependency relationship mapping diagram between entities is established;

[0103] Based on the dependency mapping graph between entities, the dependency relationships of named entities are generated.

[0104] In this embodiment, firstly, by combining the business field classification of the named entity and its position information in the text of the letter of credit terms, the semantic structure inside the text is deeply analyzed, and not only the named entity itself can be identified, but also the entity with semantic association with the named entity can be further judged, such as the subject-predicate association relationship, the verb-object association relationship, the attributive-subject association relationship, etc. Then, by analyzing the context of the named entity, the associated entity is clarified, and the information of the associated entity with the dependency relationship with the current named entity is obtained, such as identifying the position of the associated entity in the text of the letter of credit terms, and obtaining the start and end positions of the associated entity, that is, the position information of the associated entity in the text of the letter of credit terms. Finally, based on the position information of the named entity in the text of the letter of credit terms and the position information of the associated entity in the text of the letter of credit terms, a dependency mapping diagram between entities is established. The dependency mapping diagram can intuitively display the dependency relationship between the named entity and its associated entities, such as the subject-predicate, verb-object, attributive, etc., and by analyzing the mapping diagram, it is easier to identify the possible risk points or non-compliance in the terms, such as the unclear responsible party and the vague payment conditions.

[0105] In the above embodiment, by deeply analyzing the semantic association between named entities and associated entities and building a dependency mapping diagram in combination with location information, the complex semantic structure in the terms of the letter of credit is accurately captured.

[0106] Furthermore, a neural network model is used to extract and learn features of named entities and dependencies to learn key information features in the terms of the letter of credit, including:

[0107] Combine the named entities and the dependency relationships to obtain a named entity-dependency relationship combination;

[0108] Use the preset word vector model to vectorize the named entity-dependency combination to obtain the combined word vector;

[0109] Use the neural network model to learn the features of the combined word vectors, and iteratively update the neural network model based on the preset optimization algorithm until the model is fitted;

[0110] After the model is fitted, the feature learning results output by the neural network model are obtained to obtain key information features.

[0111] In this embodiment, first, the system combines the named entities and the dependencies to obtain a named entity-dependency combination. Then, the named entity and the dependency are combined to convert the named entity-dependency combination into a corresponding vector representation using a word vector model so that the neural network model can perform feature learning. The neural network model can adopt a CNN network structure or an RNN network structure. If a CNN network structure is adopted, an appropriate number of convolutional layers and neurons are designed to capture local features in the text; if an RNN network structure is adopted, an appropriate number of recursive layers and neurons are designed to capture sequence features in the text.

[0112] According to the selection of network structure, the output of the intermediate layer is transformed nonlinearly through the activation function, and the appropriate activation function is selected to enhance the model's expressiveness. The loss function is used to define the optimization target of the model, and the model performance is judged by calculating the error between the predicted output and the actual annotation. Combined with the optimization algorithm, a suitable parameter update strategy is selected for model training, and the model parameters are adjusted through algorithms such as SGD or Adam. According to the model performance during the training process, the number of training iterations and batch size are determined, and gradually adjusted to achieve the expected training effect. After the model is trained to fit the model, the learning results of the named entity and dependency features are obtained, and finally the key information features in the terms of the letter of credit are extracted.

[0113] In a specific embodiment of the present application, considering that the text of the terms of the letter of credit has certain local features and sequence features, a neural network structure combining CNN and Bi LSTM is adopted, wherein the CNN part uses 3 layers of convolution, and the convolution kernel sizes are 2*2, 3*3, and 4*4 respectively, and the number of convolution kernels in each layer is 128, and the Bi LSTM part uses 2 layers of bidirectional LSTM, and the hidden layer dimension is 256. The middle layer uses the ReLU activation function for nonlinear transformation to enhance the expressive power of the model. The model optimization goal is to minimize the cross entropy loss function, and the model parameters are updated at a preset learning rate using the Adam optimization algorithm. During the training process, the batch size is 64 for training. After each batch of training, the model performance is evaluated on the validation set. If the model performance corresponding to the training results of 5 consecutive batches is not improved, the training is terminated early. The F1 value of the final model on the test set reached 95%, which can effectively extract the key information features in the terms of the letter of credit.

[0114] In the above embodiment, deep feature extraction and learning of named entities and dependency relationships are performed through a neural network model to accurately capture key information features in the terms of the letter of credit.

[0115] Furthermore, the key information features include several groups of named entity-dependency combinations, and the steps of constructing a compliance judgment knowledge graph of the letter of credit terms and mapping the key information features to entities and relationships in the compliance judgment knowledge graph specifically include:

[0116] Cluster the business domains to which the named entities belong and obtain domain cluster labels;

[0117] Determine the letter of credit review template that matches the letter of credit clause text based on the domain clustering labels;

[0118] Constructing a compliance judgment knowledge graph of the letter of credit terms based on the letter of credit review template, wherein the compliance judgment knowledge graph includes nodes and edges;

[0119] The named entities are matched to the nodes of the compliance judgment knowledge graph according to the ontology matching algorithm, and the dependency relationships are mapped to the edges of the compliance judgment knowledge graph according to the relationship mapping rules.

[0120] In this embodiment, first, the system clusters the business domain to which the named entity belongs, obtains the domain clustering label, and determines the letter of credit review template that matches the letter of credit clause text according to the domain clustering label, and the letter of credit review template contains the business rules and laws and regulations corresponding to the letter of credit clause. Then, the letter of credit clause compliance judgment knowledge graph is constructed, and the domain ontology model is established by the ontology-based knowledge representation method. According to the business rules and laws and regulations corresponding to the letter of credit clause, the entity class, attribute and relationship are defined. Then, for several groups of named entity-dependency relationship combinations in the key information features, the ontology matching algorithm is used to map the named entity-dependency relationship combination to the ontology model, and the corresponding relationship between the named entity and the graph node is determined by the semantic similarity calculation, and the node to which the named entity belongs is determined. Then, the relationship mapping rule is used to map the dependency relationship to the predefined edge in the ontology model, and finally the compliance judgment knowledge graph corresponding to the letter of credit clause text is obtained.

[0121] In the above embodiment, by constructing a knowledge graph for compliance judgment of letter of credit terms, accurate mapping of key information features and audit rules is achieved, and the comprehensiveness and accuracy of the audit are ensured through clustering, template matching and ontology matching algorithms.

[0122] Furthermore, a review rule engine for the letter of credit is constructed, and the steps of judging whether the terms of the letter of credit meet the compliance requirements according to the review rule engine and the compliance judgment knowledge graph specifically include:

[0123] Obtain the review rules in the letter of credit review template, and use the decision tree algorithm to establish a rule model to obtain the review rule engine;

[0124] Extract key features of the letter of credit clause text to obtain key features of the text;

[0125] Use the audit rule engine to conduct compliance audit on key features of the text based on the random forest algorithm to obtain the initial non-compliant items in the text of the letter of credit terms;

[0126] Using knowledge graph reasoning technology, multi-hop reasoning is performed on the initial non-compliant items in the compliance judgment knowledge graph to obtain multi-hop reasoning results;

[0127] Based on the multi-hop reasoning results, the probability that the initial non-compliant items of the letter of credit meet the compliance requirements is calculated, and the non-compliant items are screened using the preset probability threshold to obtain the non-compliant items of the letter of credit terms.

[0128] In this embodiment, first, the system obtains the audit rules in the letter of credit audit template, and uses the decision tree algorithm to establish a rule model to obtain an audit rule engine. Then, key features are extracted according to the content of the letter of credit terms, and the extracted features are preprocessed and standardized through feature engineering to obtain text key features. Using the audit rule engine, the text key features are classified based on the random forest algorithm to determine whether the letter of credit terms meet the basic compliance requirements. When there are items that do not meet the basic compliance requirements, the initial non-compliant items in the letter of credit terms text are obtained. Finally, using the knowledge graph reasoning technology, multi-hop reasoning is performed in the compliance judgment knowledge graph to obtain a more comprehensive compliance judgment result. Based on the results of multi-hop reasoning, the Bayesian network model is used to calculate the probability that the letter of credit terms meet various compliance requirements. Through probability threshold screening, the final compliance judgment result, that is, the non-compliant items of the letter of credit terms, is determined.

[0129] In a specific embodiment of the present application, when constructing a letter of credit review rule engine, a decision tree algorithm is first used to establish a rule model. By analyzing the text data of the letter of credit terms, 20 key features are extracted, such as the letter of credit amount, beneficiary information, expiration date, etc. The features are preprocessed using the minimum-maximum normalization method, and the feature values ​​are mapped to the [0,1] interval to obtain a normalized text key feature vector. Then, the letter of credit review rule engine is used to determine whether the letter of credit terms meet the basic compliance requirements based on the decision tree classification. For the terms that do not meet the basic compliance requirements (i.e., the initial non-compliant items), the April algorithm is used to mine association rules from the compliance judgment knowledge graph, the minimum support is set to 0.5, the minimum confidence is set to 8, and 6 jump reasonings are performed in the compliance judgment knowledge graph to further judge the compliance of the initial non-compliant items, and the reasoning accuracy reaches 92%. Finally, the Bayesian network model is used to calculate the probability that the terms meet various compliance requirements, the probability threshold is set to 0.9, and the final compliance judgment results are screened. For terms and conditions that meet compliance requirements, the GPT-3 model is used to automatically generate detailed compliance reports. For terms and conditions that do not meet the requirements, a prompt message indicating that the letter of credit terms and conditions have not passed the review is output.

[0130] In the above embodiment, by constructing an audit rule engine and a compliance knowledge graph, and combining random forest and knowledge graph reasoning, an intelligent compliance review of letter of credit terms is achieved, which not only improves the efficiency of letter of credit review, but also enhances the accuracy and comprehensiveness of the review, and effectively reduces non-compliance risks.

[0131] In the above embodiment, the present application discloses a method for reviewing letter of credit terms, which belongs to the field of artificial intelligence technology. First, the text of the letter of credit terms is parsed by natural language processing technology. Then, the domain dictionary is used to perform association analysis of named entities to clarify the dependencies between entities. Then, the key information features in these named entities and their dependencies are deeply extracted and learned. A compliance judgment knowledge graph is constructed, and the extracted key information features are mapped to the corresponding entities and relationships in the graph to form a structured representation of the letter of credit terms. Finally, the constructed review rule engine is combined with the compliance judgment knowledge graph to evaluate the compliance of the letter of credit terms. Through automated processing and intelligent analysis, the present application significantly improves the efficiency and accuracy of letter of credit review, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results.

[0132] In this embodiment, the electronic device (eg Figure 1 The server shown in the figure) can receive instructions or obtain data through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0133] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned letter of credit terms text, the above-mentioned letter of credit terms text can also be stored in a blockchain node.

[0134] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0135] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the computer-readable instructions are executed, they can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0136] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0137] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a device for reviewing letter of credit terms. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0138] like Figure 3 As shown, the credit terms review device 300 described in this embodiment includes:

[0139] The text processing module 301 is used to obtain the text of the letter of credit terms and perform natural language processing on the text of the letter of credit terms, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition;

[0140] The association analysis module 302 is used to perform association analysis on the named entities in the text of the letter of credit terms according to the pre-established domain dictionary to obtain the dependency relationship of the named entities;

[0141] A feature learning module 303 is used to extract and learn features of named entities and dependency relationships using a neural network model to learn key information features in the terms of the letter of credit;

[0142] A graph construction module 304 is used to construct a compliance judgment knowledge graph of the letter of credit terms and map key information features to entities and relationships of the compliance judgment knowledge graph;

[0143] The terms review module 305 is used to build a review rule engine for the letter of credit, and judge whether the terms of the letter of credit meet the compliance requirements based on the review rule engine and the compliance judgment knowledge graph.

[0144] Furthermore, the text processing module 301 is specifically used for:

[0145] The text of the letter of credit terms is obtained, and data structure processing is performed on the text of the letter of credit terms to obtain structured text data; word segmentation operation is performed on the structured text data to obtain text segmentation; part-of-speech tagging is performed on the text segmentation to identify the part-of-speech information of the text segmentation; based on the part-of-speech information of the text segmentation, named entity recognition technology is used to identify and extract named entities in the text of the letter of credit terms.

[0146] Furthermore, the association analysis module 302 is specifically used for:

[0147] According to the domain category defined in the domain dictionary, the business domain to which the named entity belongs is determined; the entity start position and entity end position of the named entity are identified to obtain the position information of the named entity in the text of the letter of credit terms; based on the business domain to which the named entity belongs and the position information of the named entity in the text of the letter of credit terms, the named entity is associated with the analysis to obtain the dependency relationship of the named entity.

[0148] Furthermore, the association analysis module 302 is further used to:

[0149] Determine the associated entity that has semantic association with the named entity according to the business field to which the named entity belongs and the position information of the named entity in the text of the letter of credit terms, wherein the semantic association includes subject-predicate association relationship, verb-object association relationship, and attributive-predicate association relationship; identify the entity start position and entity end position of the associated entity, and obtain the position information of the associated entity in the text of the letter of credit terms;

[0150] According to the position information of the named entity in the text of the letter of credit terms and the position information of the associated entity in the text of the letter of credit terms, a dependency mapping diagram between entities is established; based on the dependency mapping diagram between entities, the dependency relationship of the named entity is generated.

[0151] Furthermore, the feature learning module 303 is specifically used for:

[0152] Combine named entities and dependency relationships to obtain a named entity-dependency combination; use a preset word vector model to vectorize the named entity-dependency combination to obtain a combined word vector; use a neural network model to learn features of the combined word vector, and iteratively update the neural network model based on a preset optimization algorithm until the model is fitted; when the model is fitted, obtain the feature learning results output by the neural network model to obtain key information features.

[0153] Furthermore, the key information features include several groups of named entity-dependency combinations, and the graph construction module 304 is specifically used to:

[0154] Cluster the business domains to which the named entities belong to obtain domain clustering labels; determine the letter of credit review template that matches the text of the letter of credit terms based on the domain clustering labels; construct a compliance judgment knowledge graph of the letter of credit terms based on the letter of credit review template, wherein the compliance judgment knowledge graph includes nodes and edges; match the named entities to the nodes of the compliance judgment knowledge graph based on the ontology matching algorithm, and map the dependency relationships to the edges of the compliance judgment knowledge graph based on the relationship mapping rules.

[0155] Furthermore, the clause review module 305 is specifically used for:

[0156] The audit rules in the letter of credit audit template are obtained, and the decision tree algorithm is used to establish a rule model to obtain the audit rule engine; the key features of the letter of credit terms text are extracted to obtain the key features of the text; the audit rule engine is used to perform compliance audit on the key features of the text based on the random forest algorithm to obtain the initial non-compliant items in the letter of credit terms text; the knowledge graph reasoning technology is used to perform multi-hop reasoning on the initial non-compliant items in the compliance judgment knowledge graph to obtain the multi-hop reasoning results; based on the multi-hop reasoning results, the probability that the initial non-compliant items of the letter meet the compliance requirements is calculated, and the non-compliant items are screened through the preset probability threshold to obtain the non-compliant items of the letter of credit terms.

[0157] In the above embodiment, the present application discloses a device for reviewing letter of credit terms, which belongs to the field of artificial intelligence technology. First, the text of the letter of credit terms is parsed by natural language processing technology. Then, the domain dictionary is used to perform association analysis of named entities to clarify the dependencies between entities. Then, the key information features in these named entities and their dependencies are deeply extracted and learned. A compliance judgment knowledge graph is constructed, and the extracted key information features are mapped to the corresponding entities and relationships in the graph to form a structured representation of the letter of credit terms. Finally, the constructed review rule engine is combined with the compliance judgment knowledge graph to evaluate the compliance of the letter of credit terms. Through automated processing and intelligent analysis, the present application significantly improves the efficiency and accuracy of letter of credit review, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results.

[0158] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0159] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 having a memory 41, a processor 42, and a network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital processor (DSP), an embedded device, etc.

[0160] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0161] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the credit terms review method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0162] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as running the computer-readable instructions of the credit terms review method.

[0163] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0164] In the above embodiment, the present application discloses a computer device, which belongs to the field of artificial intelligence technology. First, the text of the letter of credit terms is parsed by natural language processing technology. Then, the domain dictionary is used to perform association analysis of named entities to clarify the dependencies between entities. Then, the key information features in these named entities and their dependencies are deeply extracted and learned. A compliance judgment knowledge graph is constructed, and the extracted key information features are mapped to the corresponding entities and relationships in the graph to form a structured representation of the letter of credit terms. Finally, the compliance of the letter of credit terms is evaluated by combining the compliance judgment knowledge graph through the constructed audit rule engine. Through automated processing and intelligent analysis, this application significantly improves the efficiency and accuracy of letter of credit review, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results.

[0165] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the letter of credit terms review method as described above.

[0166] In the above embodiment, the present application discloses a computer-readable storage medium, which belongs to the field of artificial intelligence technology. First, the text of the letter of credit terms is parsed by natural language processing technology. Then, the domain dictionary is used to perform association analysis of named entities to clarify the dependencies between entities. Then, the key information features in these named entities and their dependencies are deeply extracted and learned. A compliance judgment knowledge graph is constructed, and the extracted key information features are mapped to the corresponding entities and relationships in the graph to form a structured representation of the letter of credit terms. Finally, the compliance of the letter of credit terms is evaluated by combining the compliance judgment knowledge graph through the constructed audit rule engine. Through automated processing and intelligent analysis, this application significantly improves the efficiency and accuracy of letter of credit review, reduces the error rate, ensures the consistency of letter of credit review standards, and not only reduces the manual burden, but also improves the reliability and transparency of letter of credit review results.

[0167] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0168] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0169] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for reviewing letter of credit terms, characterized in that: include: Obtaining a text of the terms of the letter of credit, and performing natural language processing on the text of the terms of the letter of credit, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition; According to a pre-established domain dictionary, association analysis is performed on the named entities in the text of the letter of credit terms to obtain dependency relationships of the named entities; Using a neural network model, extracting and learning features of the named entities and the dependency relationships, so as to learn key information features in the terms of the letter of credit; Constructing a compliance judgment knowledge graph of the letter of credit terms, and mapping the key information features to entities and relationships of the compliance judgment knowledge graph; Construct a review rule engine for the letter of credit, and determine whether the terms of the letter of credit meet the compliance requirements based on the review rule engine and the compliance judgment knowledge graph.

2. The method for reviewing letter of credit terms as claimed in claim 1, characterized in that: The step of obtaining the text of the letter of credit terms and performing natural language processing on the text of the letter of credit terms specifically includes: Acquiring the text of the letter of credit terms and conditions, and performing data structure processing on the text of the letter of credit terms and conditions to obtain structured text data; Performing a word segmentation operation on the structured text data to obtain text segmentation; Performing part-of-speech tagging on the text segmentation to identify part-of-speech information of the text segmentation; According to the part-of-speech information of the text segmentation, named entity recognition technology is used to identify and extract the named entities in the text of the letter of credit terms.

3. The method for reviewing letter of credit terms as claimed in claim 1, characterized in that: The step of performing association analysis on the named entities in the text of the letter of credit terms according to the pre-established domain dictionary to obtain the dependency relationship of the named entities specifically includes: Determining the business domain to which the named entity belongs according to the domain category defined in the domain dictionary; Identify the entity start position and entity end position of the named entity to obtain the position information of the named entity in the text of the letter of credit terms; Based on the business field to which the named entity belongs and the position information of the named entity in the text of the letter of credit terms, the named entity is subjected to association analysis to obtain the dependency relationship of the named entity.

4. The method for reviewing letter of credit terms as claimed in claim 3, characterized in that: The step of performing association analysis on the named entities based on the business fields to which the named entities belong and the position information of the named entities in the text of the letter of credit terms to obtain the dependency relationship of the named entities specifically includes: Determine, according to the business field to which the named entity belongs and the position information of the named entity in the text of the letter of credit terms, an associated entity having a semantic association with the named entity, wherein the semantic association includes a subject-predicate association relationship, a verb-object association relationship, and a subject-predicate association relationship; Identify the entity start position and entity end position of the associated entity, and obtain the position information of the associated entity in the text of the letter of credit terms; Establishing a dependency relationship mapping diagram between entities according to the position information of the named entity in the text of the letter of credit terms and the position information of the associated entity in the text of the letter of credit terms; Based on the dependency relationship mapping graph between the entities, the dependency relationships of the named entities are generated.

5. The method for reviewing letter of credit terms as claimed in claim 3, characterized in that: The step of using a neural network model to extract and learn features of the named entities and the dependency relationships to learn key information features in the terms of the letter of credit specifically includes: Combining the named entity and the dependency relationship to obtain a named entity-dependency relationship combination; Using a preset word vector model to perform word vectorization on the named entity-dependency combination to obtain a combined word vector; Using the neural network model to perform feature learning on the combined word vector, and iteratively updating the neural network model based on a preset optimization algorithm until the model is fitted; After the model is fitted, the feature learning results output by the neural network model are obtained to obtain the key information features.

6. The method for reviewing letter of credit terms as claimed in claim 5, characterized in that: The key information features include a plurality of named entity-dependency combinations. The steps of constructing the compliance judgment knowledge graph of the letter of credit terms and mapping the key information features to entities and relationships in the compliance judgment knowledge graph specifically include: Clustering the business fields to which the named entities belong to obtain field cluster labels; Determine a letter of credit review template matching the letter of credit clause text according to the domain clustering label; Constructing a compliance judgment knowledge graph of the letter of credit terms according to the letter of credit review template, wherein the compliance judgment knowledge graph includes nodes and edges; The named entities are matched to the nodes of the compliance judgment knowledge graph according to the ontology matching algorithm, and the dependency relationships are mapped to the edges of the compliance judgment knowledge graph according to the relationship mapping rules.

7. The method for reviewing letter of credit terms as claimed in claim 6, characterized in that: The step of constructing a review rule engine for a letter of credit and judging whether the terms of the letter of credit meet the compliance requirements according to the review rule engine and the compliance judgment knowledge graph specifically includes: Acquire the audit rules in the letter of credit audit template, and establish a rule model using a decision tree algorithm to obtain the audit rule engine; Extract key features of the letter of credit clause text to obtain key features of the text; Using the audit rule engine, performing compliance audit on the key features of the text based on a random forest algorithm, and obtaining initial non-compliant items in the text of the letter of credit terms; Using knowledge graph reasoning technology, multi-hop reasoning of the initial non-compliant items is performed in the compliance judgment knowledge graph to obtain a multi-hop reasoning result; Based on the multi-hop reasoning result, the probability that the initial non-compliant items of the letter of credit meet the compliance requirements is calculated, and the non-compliant items are screened using a preset probability threshold to obtain the non-compliant items of the letter of credit terms.

8. A device for reviewing letter of credit terms, characterized in that: include: A text processing module, used to obtain the text of the letter of credit terms and conditions, and perform natural language processing on the text of the letter of credit terms and conditions, wherein the natural language processing includes word segmentation, part-of-speech tagging, and named entity recognition; An association analysis module, used to perform association analysis on the named entities in the text of the letter of credit terms according to a pre-established domain dictionary, and obtain dependency relationships of the named entities; A feature learning module, used for extracting and learning features of the named entities and the dependency relationships using a neural network model, so as to learn key information features in the terms of the letter of credit; A graph construction module, used to construct a compliance judgment knowledge graph of the letter of credit terms, and map the key information features to entities and relationships of the compliance judgment knowledge graph; The terms review module is used to build a review rule engine for the letter of credit, and judge whether the terms of the letter of credit meet the compliance requirements based on the review rule engine and the compliance judgment knowledge graph.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the method for reviewing letter of credit terms as claimed in any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the method for reviewing letter of credit terms as claimed in any one of claims 1 to 7 are implemented.

Images