Electronic contract risk identification method and device, equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, discloses an electronic contract risk identification method and device, electronic equipment, a computer readable storage medium and a computer program product, and is used for solving the problem that a risk identification method in the prior art cannot comprehensively capture the signing risk of an electronic contract. And the risk identification sensitivity and accuracy are not enough. The method comprises the following steps: acquiring an electronic contract text and signing behavior data generated when a user signs an electronic contract; signature behavior sequence features of the signature behavior data are extracted, and text content features of the electronic contract text are obtained; and calling a risk judgment model, and performing risk identification on the electronic contract based on the signature behavior sequence features and the text content features. According to the method, the text of the electronic contract and the risk information in the signing process can be comprehensively considered, so that the signing risk of the electronic contract is comprehensively captured, and the risk identification sensitivity and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an electronic contract risk identification method, apparatus, electronic device, computer storage medium, and computer program product. Background Technology

[0002] With the rapid development of internet finance and digital office practices, electronic contracts have become a widely adopted form of signing and legal proof for individuals and businesses. The electronic contract signing process generally employs digital signatures, identity verification, and encryption technologies to ensure the authenticity of the signing parties and the integrity of the contract content. However, these traditional security measures primarily focus on static identity authentication and data integrity checks, making it difficult to effectively identify abnormal behavior during the signing process or fraud risks hidden in the contract terms. For example, a signatory may use a proxy, automated scripts, or false supporting documents to complete the signing, evading genuine identity verification and making the risk difficult to detect.

[0003] Existing technologies employ various methods to monitor signing processes, such as handwritten signature pads, cameras, or biometric hardware. However, these methods often rely on specific hardware devices, resulting in high costs and susceptibility to limitations imposed by the signing environment. Furthermore, forged signatures and non-compliant contract terms typically occur simultaneously, suggesting a correlation between them. Existing technologies do not consider the interrelationships between these various risk characteristics. Therefore, we propose a technological solution capable of comprehensively identifying and assessing risks associated with electronic contracts throughout the entire process. Summary of the Invention

[0004] The main objective of this invention is to solve the technical problem that the existing technology cannot fully capture the risks of signing electronic contracts, resulting in insufficient sensitivity and accuracy in risk identification.

[0005] The first aspect of this invention provides a method for identifying risks in electronic contracts, comprising: Obtain electronic contract texts and user signing behavior data generated when signing electronic contracts; Extract the signing behavior sequence features from the signing behavior data, and obtain the text content features of the electronic contract text; The risk assessment model is invoked to identify risks in the electronic contract based on the characteristics of the signing behavior sequence and the characteristics of the text content.

[0006] Optionally, in a first implementation of the first aspect of the present invention, after extracting the signing behavior sequence features of the signing behavior data and obtaining the text content features of the electronic contract text, the method further includes: The multi-dimensional temporal action feature extraction model is invoked to extract the signature behavior sequence feature vector contained in the signature behavior data; The textual risk feature vector of the electronic contract text is extracted by calling a natural language processing model; The risk assessment model, based on the signature behavior sequence characteristics and the text content characteristics, identifies risks in the electronic contract, including: The signature behavior sequence feature vector and the text risk feature vector are cross-fused to obtain fused vector information; The risk assessment model is invoked to identify risks in the electronic contract based on the fused vector information, and the risk identification results are obtained.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the risk assessment model is a multi-head attention structure; The step of cross-fusing the signature behavior sequence feature vector and the text risk feature vector to obtain the fused vector information includes: Calculate the similarity information between the signature behavior sequence feature vector and the text risk feature vector, respectively; The query dot product matrix is ​​obtained based on the similarity information, and an attention matrix is ​​constructed based on the query dot product matrix. Calculate the attention weights based on the attention matrix; Based on the multi-head attention structure and the attention weights, the signature behavior sequence feature vector and the text risk feature vector are weighted and cross-fused to obtain fused vector information.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the step of invoking the risk assessment model to identify risks in the electronic contract based on the fused vector information, and obtaining the risk identification result, includes: The risk assessment model is invoked to perform risk assessment on the fused vector information based on the attention matrix, and a risk assessment score is output.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, after extracting the signing behavior sequence features of the signing behavior data and obtaining the text content features of the electronic contract text, the method further includes: The signing behavior sequence features and the text content features are mapped to graph nodes to construct a heterogeneous graph; Risk association information is obtained by performing information aggregation and multi-layer learning on information in the heterogeneous graphs based on graph convolution or graph attention networks. The risk assessment model, based on the signature behavior sequence characteristics and the text content characteristics, identifies risks in the electronic contract, including: The risk assessment model is invoked to identify risks in the electronic contract based on the risk association information.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the method further includes, after performing risk identification on the electronic contract: Obtain the manual review results for data deemed to contain risks, and annotate the data based on the review results; The risk assessment model is optimized and updated based on the labeled verification data.

[0011] A second aspect of the present invention provides an electronic contract risk identification device, comprising: The information acquisition module is used to acquire the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; The feature extraction module is used to extract the signing behavior sequence features of the signing behavior data and obtain the text content features of the electronic contract text. The risk identification module is used to invoke the risk assessment model to identify risks in the electronic contract based on the signature behavior sequence characteristics and the text content characteristics.

[0012] A third aspect of the present invention provides an electronic contract risk identification device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the electronic contract risk identification device to perform the steps of the above-described electronic contract risk identification method.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the steps of the above-described electronic contract risk identification method.

[0014] A fifth aspect of the present invention provides a computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the electronic contract risk identification method described above.

[0015] The technical solution provided by this invention involves acquiring electronic contract text and signing behavior data generated by users when signing electronic contracts; extracting the signing behavior sequence features from the signing behavior data and acquiring the text content features of the electronic contract text; and calling a risk assessment model to identify risks in the electronic contract based on the signing behavior sequence features and text content features. This method can comprehensively consider the text of the electronic contract and the risk information during the signing process, thereby comprehensively capturing the signing risks of electronic contracts and improving the sensitivity and accuracy of risk identification. Furthermore, the device, electronic device, computer-readable storage medium, and computer program product provided by this invention also solve the corresponding technical problems. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the first embodiment of the electronic contract risk identification method in this invention. Figure 2 This is a flowchart illustrating the second embodiment of the electronic contract risk identification method in this invention. Figure 3 This is a flowchart illustrating the third embodiment of the electronic contract risk identification method in this invention. Figure 4 This is a schematic diagram of one embodiment of the electronic contract risk identification device in this invention; Figure 5 This is a schematic diagram of one embodiment of the electronic contract risk identification device in this invention; Figure 6 This is a schematic diagram illustrating the principle of a computer-readable medium according to an embodiment of the present invention. Detailed Implementation

[0017] Exemplary embodiments of the invention will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limiting the invention to the embodiments set forth herein. Rather, these exemplary embodiments are provided to make the invention more comprehensive and complete, and to facilitate a full communication of the inventive concept to those skilled in the art. The same reference numerals in the drawings denote the same or similar elements, components, or parts, and therefore repeated descriptions of them will be omitted.

[0018] Subject to the technical concept of this invention, the features, structures, characteristics or other details described in a particular embodiment may be combined in one or more other embodiments in a suitable manner.

[0019] In the description of specific embodiments, the features, structures, characteristics, or other details described in this invention are intended to enable those skilled in the art to fully understand the embodiments. However, it is not excluded that those skilled in the art can practice the technical solutions of this invention without one or more of the specific features, structures, characteristics, or other details.

[0020] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0021] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0022] The terms “and / or” or “and / or” include all combinations of any one or more of the listed items.

[0023] Please see Figure 1 The first embodiment of the electronic contract risk identification method in this invention includes: S101. Obtain the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; It is understood that the executing entity of this invention can be an electronic contract risk identification device, a terminal, or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0024] Since this embodiment considers not only the risks in the electronic contract text but also the risks in the signing process when identifying and assessing electronic contract risks, the server first needs to obtain the electronic contract text and the signing behavior data generated by the user when signing the electronic contract after receiving the electronic contract risk identification request.

[0025] The acquisition of the electronic contract text is relatively easy to understand, such as segmenting the text and extracting risk keywords, which will not be elaborated here; however, the acquisition of signing behavior data generated by the user when signing the electronic contract can be signing action data during the electronic contract signing process, which can include information such as the handwriting trajectory, pressure, operation speed, acceleration, and timestamp when the user signs through a touch screen, stylus, mouse or other input device.

[0026] S102. Extract the signing behavior sequence features from the signing behavior data and obtain the text content features of the electronic contract text; Once the signing behavior data and electronic contract text are obtained, the features contained in this information can be extracted so that contract risks can be identified and judged based on these features.

[0027] In one specific implementation, obtaining the signing behavior sequence features of the signing behavior data can be achieved by calling a multi-dimensional temporal action feature extraction model to extract the signing behavior sequence feature vector contained in the signing behavior data; using a trained artificial intelligence signing behavior analysis model, feature extraction and pattern recognition are performed on the multi-dimensional temporal features included in the signing action data, such as handwriting trajectory, pressure, operation speed, acceleration, and timestamps when signing via touch screen, stylus, mouse, or other input devices, to obtain the signing behavior sequence features. Obtaining the text content features of the electronic contract text can be achieved by calling a natural language processing model to extract the text risk feature vector of the electronic contract text. The natural language processing model can perform semantic and structural analysis on the electronic contract text content, including segmenting the contract text into paragraphs and sentences, and analyzing the content of each clause to identify possible ambiguities, incompleteness, or implicit clauses, thereby obtaining the text content features; in a preferred implementation, the text can be used to extract a sensitivity distribution vector through a pre-trained language model or rule engine.

[0028] S103. Invoke the risk assessment model to identify risks in electronic contracts based on the characteristics of the signing behavior sequence and the characteristics of the text content.

[0029] In this embodiment, before performing specific risk identification and assessment, a comprehensive risk judgment model is constructed. Based on this model, the results of signing behavior analysis and contract text analysis are combined to conduct a comprehensive risk assessment. Specifically, the risk judgment model can be a classification model built based on artificial intelligence neural networks. Electronic contract data labeled with risk information is used to construct training, testing, and validation sets. Based on the constructed training and testing sets, the classification model is optimized and trained, and the hyperparameters in the model are adjusted to obtain the final risk judgment model capable of risk identification.

[0030] In this step, the signing behavior sequence features and text content features obtained in the previous steps are input into the risk assessment model for risk identification. Based on the risk assessment model, it analyzes whether the signing order is abnormal, whether the signing duration and frequency change abruptly, whether the signing environment (such as IP address, device information, geographical location, etc.) has changed significantly, whether there is a significant difference between the current operation and the user's historical behavior profile, and whether there is unreasonable information in the contract text. In this embodiment, the risk assessment model integrates the analysis results of the signing behavior sequence features and the analysis results of the text content features to give a comprehensive risk score to the current signing process. Risk assessment can be calculated using a rule engine or a machine learning model. Example rules include: if the signing behavior is marked as abnormal in the analysis and the contract text contains at least one high-risk clause, the risk score is increased; if the user has no history of proxy signing and the behavior pattern is close to normal, the risk score is decreased. Simultaneously, the signing order is also considered. For example, if the order of signing the contract by both parties does not conform to the normal process (e.g., Party A should sign first but Party B signs first), it can also be considered abnormal.

[0031] In one specific implementation, when using a risk assessment model for risk identification, a weighted calculation method can be used to comprehensively consider the characteristics of the signing behavior sequence and the characteristics of the text content.

[0032] If the risk assessment result exceeds a preset threshold in this step, the system will trigger an anomaly warning. After triggering the warning, the system will activate the warning and manual review module. For example, the risk assessment model described in this embodiment can generate contract anomaly markers when outputting the risk identification results. For instance, when the contract text contains common risk terms or high-risk expressions, such as keywords like "disclaimer," "unlimited," or "not specified," the system will determine whether there is any ambiguity or lack of explanation in the context. For example, if the contract contains a statement like "Party A shall bear all responsibilities" without specific limitations on the scope of responsibility, the system will consider this clause to have a high potential risk and mark it accordingly.

[0033] Furthermore, after risk identification and risk scoring, real-time notifications are generated based on the scoring results, displaying detailed anomaly information on the management interface or security personnel's workstations. For example, the review interface displays a behavioral graph of the signing process and triggering points for anomalies (such as sudden spikes in the signing speed curve), along with possible causes. Simultaneously, clauses analyzed as high-risk are highlighted on the contract text interface, along with risk descriptions. Reviewers can view this information and compare it with reference data on normal user behavior (such as handwriting comparisons from past signatures by the same user) to determine if fraud or errors occurred during the signing. If reviewers confirm an anomaly, they can halt the signing process and conduct further investigation; if it's a false alarm, the review results are recorded and fed back to the system. Reviewers' feedback and processing results are recorded in logs for model training.

[0034] The method provided in this embodiment of the invention can comprehensively consider the text of the electronic contract and the risk information in the signing process, thereby fully capturing the signing risks of electronic contracts and improving the sensitivity and accuracy of risk identification.

[0035] Please refer to Figure 2 The second embodiment of the electronic contract signing risk identification method in this invention includes: S201. Obtain the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; S202. Extract the signing behavior sequence features from the signing behavior data and obtain the text content features of the electronic contract text; The contents of steps S201-S202 in this embodiment are basically the same as those of steps S101-S102 in the previous embodiment, so they will not be repeated here.

[0036] S203. Use the multi-dimensional temporal action feature extraction model to extract the signature behavior sequence feature vector contained in the signature behavior data; In this embodiment, the multi-dimensional temporal action feature extraction model can be a pre-trained model based on artificial intelligence algorithms. It can analyze signing behavior and extract the signing behavior sequence feature vector contained in the signing behavior data. The signing behavior data may include multi-dimensional temporal features such as handwriting trajectory, pressure, operation speed, acceleration, and timestamps when a user signs using a touchscreen, stylus, mouse, or other input devices. The multi-dimensional temporal action feature extraction model performs feature extraction and pattern recognition on these features to obtain the signing behavior sequence feature vector. Subsequently, based on the signing behavior sequence feature vector, it can be determined whether the current signing behavior is consistent with the historical signing behavior characteristics of the contract signatory, and whether there are abnormal behaviors such as unauthorized signing or automated script operations.

[0037] S204. Use a natural language processing model to extract the textual risk feature vector of the electronic contract text; The natural language processing model can be a pre-built model based on Natural Language Processing (NLP) algorithms. It can perform semantic and structural analysis on the electronic contract text, including segmenting and dividing the contract text into sentences, and analyzing the content of each clause to identify potential ambiguities, incompleteness, or implicit clauses. For example, it can detect the presence of vague disclaimers, unusually high-risk clauses, unexplained matters, and whether important protective clauses are missing. The system automatically marks and prompts potential risk points, providing a basis for subsequent risk assessment. This module automates and intelligently reviews contract texts, quickly identifying contract loopholes that are easily overlooked by traditional manual review.

[0038] S205. Cross-fuse the signature behavior sequence feature vector and the text risk feature vector to obtain fused vector information; S206. Call the risk assessment model, perform risk assessment on the fused vector information based on the attention matrix, and output a risk assessment score; In this embodiment, the risk assessment model used is a multi-head attention structure. Based on this multi-head attention structure, this scheme differs from the general linear weighting scheme. This scheme achieves non-linear fusion of information through cross-fusion, associating and calculating the user's signing behavior sequence feature vector and the text risk feature vector through a multi-head attention mechanism or graph structure, thereby ultimately obtaining a context-aware comprehensive risk representation.

[0039] This embodiment uses a multi-head attention mechanism as an example for explanation. Represents the feature vector of the signing behavior sequence, in order to The text risk feature vector is represented by a multi-head cross-attention mechanism, which integrates the signature behavior sequence feature vector. and text risk feature vector The specific steps for deep interactive integration are as follows: (1) Input processing: The signature behavior sequence feature vector and text risk feature vector Projected into public spaces.

[0040] (2) Attention role allocation: Set the feature vector of signing behavior sequence As query Q, text risk feature vector As key K and value V.

[0041] (3) Calculate attention weights: (3.1) Calculate the similarity between Q and K: that is, calculate the dot product of vector Q and key vector K. Where T represents transpose; (3.2) Gradient stabilization of the similarity results yields the scaled dot product matrix. The size is n×m, where n is the number of queries and m is the number of keys.

[0042] (3.3) The attention matrix A = softmax( is obtained using the softmax function.) ).

[0043] Where A ij =softmax( )= (where 0) ij <1, and the sum of each row of matrix A is 1. A ij This represents the "degree of attention" of the i-th query vector (signature behavior sequence feature vector) to the j-th key vector (text risk feature vector). Example: If A... 23 =0.9, indicating that the second behavioral feature point pays high attention to the third text feature point, with an attention coefficient of 0.9.

[0044] (4) Final output: Context vector The C-value is mapped to the final risk assessment score through linear transformation and nonlinear activation. Unlike linear weighting, the attention mechanism allows signing features to be dynamically reweighted in the context of the text, automatically focusing on the most relevant risk paragraphs in the contract, thus achieving nonlinear interactive enhancement of features.

[0045] S207. Obtain the manual review results for data deemed to contain risks, and annotate the data based on the review results; After risk identification and risk scoring are performed, real-time notifications are generated based on the scoring results, displaying detailed anomaly information on the management interface or security personnel's workstations. For example, the review interface displays a behavioral graph of the signing process and triggering points for anomalies (such as sudden spikes in the signing speed curve), along with possible causes. Simultaneously, clauses analyzed as high-risk are highlighted on the contract text interface, along with risk descriptions. Reviewers can view this information and compare it with reference data on normal user behavior (such as handwriting comparisons from past signatures by the same user) to determine if fraud or errors occurred during the signing. If reviewers confirm an anomaly, they can halt the signing process and conduct further investigation; if it's a false alarm, the review results can be recorded and fed back to the system. This method enables manual review, and the review results are annotated on the data.

[0046] S208. Optimize and update the risk assessment model based on the labeled verification data. ​

[0047] Based on the labeled results, the feedback and processing results of the reviewers are recorded in the log for subsequent optimization and training of the risk assessment model, so as to achieve adaptive updates of the model.

[0048] The method provided in this embodiment of the invention can adaptively adjust the weight of behavioral features according to the text context based on the structure of cross-modal attention. It can also automatically amplify the influence of the corresponding sensitive clauses in the contract when the signing trajectory shows high anomaly through the attention structure. It can comprehensively consider the text of the electronic contract and the risk information in the signing process, thereby comprehensively capturing the signing risks of electronic contracts and improving the sensitivity and accuracy of risk identification.

[0049] Please refer to Figure 3 The third embodiment of the electronic contract signing risk identification method in this invention includes: S301. Obtain the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; S302. Extract the signing behavior sequence features from the signing behavior data and obtain the text content features of the electronic contract text; The contents of steps S301-S302 in this embodiment are basically the same as those of steps S101-S102 in the previous embodiment, so they will not be repeated here.

[0050] S303. Map the signing behavior sequence features and text content features to graph nodes to construct a heterogeneous graph; S304. Information aggregation and multi-layer learning are performed on information in heterogeneous graphs based on graph convolution or graph attention networks to obtain risk association information; In this embodiment, the interaction between signing behavior sequence features and text content features is realized by constructing a heterogeneous graph. The signing behavior sequence features and text content features are constructed into a heterogeneous graph, and graph neural networks (GNNs) are used to perform deep fusion of the graph structure to obtain a fusion vector or graph embedding with mixed semantics.

[0051] For example, the signing period and contract terms can be mapped to graph nodes, and edges can be set between behavior nodes and corresponding text nodes. Graph Convolutional Networks (GCNs) or Graph Attention Networks (GATs) can be used to achieve information interaction through neighbor propagation and attention aggregation. Graph networks can capture global dependencies at a higher level, allowing the impact of signing behavior on high-risk text nodes, as well as the feedback of high-risk text on abnormal behavior, to propagate throughout the graph, and learn rich representations through multi-layered networks.

[0052] To illustrate this with a specific example, the steps include the following: (1) Graph construction: Nodes: Map behavioral events such as "signing time period" and "specific signing action" to one type of node; map textual risk points such as "contract terms" and "sensitive word location" to another type of node.

[0053] Edges: Establishing connections (edges) between a behavior node and text risk nodes that it is directly associated with or affects. For example, if a signature action occurs when a high-risk clause is read, there is an edge between them.

[0054] (2) Information dissemination and aggregation (GNN core): Each node collects information about its neighboring nodes (nodes connected via edges) and aggregates this information using GCN or GAT mechanisms. GCN: Typically, neighbor information is averaged or weighted (based on graph structure).

[0055] GAT: A more advanced approach, it uses an attention mechanism to calculate the attention weight of the current node for each of its neighbors, and then performs a weighted summation of the neighbor information based on these weights. This allows nodes to selectively focus on more important neighbors.

[0056] The aggregated neighbor information is combined with the node's own information and then transformed through a neural network to update the node's representation.

[0057] (3) Multi-layer learning: Repeat the above information propagation and aggregation steps multiple times (multi-layer network). Each layer allows information to propagate further in the graph, and nodes can capture a wider range of contextual information.

[0058] After multi-layer GNN learning, the final representation of each node (whether behavioral or textual) incorporates relevant structural and neighbor information from the graph, containing richer semantic and risk-related information.

[0059] S305. Use the risk assessment model to identify risks in electronic contracts based on risk association information.

[0060] The specific technical solution in this embodiment is basically the same as the content in step S103 of the previous embodiment.

[0061] Furthermore, in a preferred embodiment, this embodiment also includes a technical solution for optimizing the scoring system of the risk assessment score. Risk indicators (behavioral anomaly degree, textual risk degree) from different modalities and with inherent uncertainties are integrated to obtain the final risk assessment score.

[0062] Its mechanism is based on a fuzzy inference system using the Mamdani or Sugeno method. A fuzzy rule engine based on the Mamdani or Sugeno method is constructed, fuzzifying the anomaly indicators of signing behavior (such as signing speed, pressure, and trajectory jitter) and the risk measures of the text (such as sensitive word count and clause complexity) into membership functions. By defining fuzzy rules such as "if the behavior has high anomaly and the text has high risk, then output extremely high risk," fuzzy inference is used to obtain a comprehensive risk membership degree, which is then defuzzified to obtain the final risk score. This multi-layered fuzzy inference structure, combined with the weight allocation of the analytic hierarchy process, can non-linearly integrate multi-dimensional indicators, significantly reducing the computational complexity of linear assessment and objectively reflecting the overall risk.

[0063] The method specifically includes the following: (1) Fuzzification: Converting clear input indicators into membership degrees of fuzzy sets. For example: Abnormal behavior indicators include: "signature speed", "pen pressure", and "trajectory jitter". For each indicator, a fuzzy set (e.g., "low", "medium", "high" anomalies) and a corresponding membership function (e.g., triangle, trapezoidal function) are defined.

[0064] Text risk measures include "number of times sensitive words appear" and "logical complexity of clauses". Similarly, fuzzy sets (such as "low risk", "medium risk", and "high risk") and membership functions are defined.

[0065] Fuzzy rule base: Defines a series of IF-THEN fuzzy rules to reflect expert knowledge.

[0066] For example: IF: high behavioral abnormality AND high text risk, THEN: extremely high overall risk; IF behavioral anomaly degree AND textual risk, THEN comprehensive risk; IF indicates low behavioral anomaly, OR indicates low textual risk, THEN indicates low overall risk.

[0067] (2) Fuzzy Inference: For a given input value (e.g., speed = slow, jitter = large, number of sensitive words = 5), calculate which rules' antecedents (i.e., the IF part in the previous example) they trigger.

[0068] Based on the logic of the conditions in the rule antecedent (AND usually takes the minimum, OR usually takes the maximum), calculate the activation strength (i.e. the credibility of the rule conclusion) of each triggered rule.

[0069] Apply activation intensity to the output fuzzy set corresponding to the conclusion of the "clipping" or "scaling" rule (THEN part).

[0070] Aggregate the output fuzzy sets of all triggered rules (usually by taking the maximum union) to form a comprehensive output fuzzy set, which represents the membership degree of the final risk to different levels (such as "low", "medium", "high", "extremely high").

[0071] (3) Defuzzification: The aggregated output fuzzy set (a fuzzy distribution) is transformed back into a clear, usable final risk score. A common method is the centroid method, which calculates the center of the area under the fuzzy distribution curve.

[0072] Fuzzy logic is naturally suited to handling vague and ambiguous concepts such as "high anomaly" and "high risk." The rule base directly encodes the experience and intuitive judgment logic of domain experts. The fuzzy reasoning process (especially Mamdani) is inherently non-linear, capturing complex interactions between indicators. Compared to training a deep neural network to fuse these indicators, rule-based fuzzy systems typically offer more controllable computational complexity during inference.

[0073] This layered fusion architecture aims to fully leverage the complementarity of different modalities of information and deepen the interaction and understanding of information layer by layer through different mechanisms (attention, graph propagation, fuzzy reasoning), ultimately achieving a more accurate, robust, and interpretable contract signing risk assessment.

[0074] The method provided in this invention utilizes graph neural networks for multi-hop information transmission, enabling a complex topological structure to be formed between signing events and textual terms through edge connections. This structure ensures fine-grained interaction across modalities, allowing for multi-layered attention between nodes in different modalities through graph attention mechanisms. The learned representations are more robust and richer, comprehensively considering both the text of the electronic contract and risk information during the signing process, thereby comprehensively capturing the signing risks of electronic contracts and improving the sensitivity and accuracy of risk identification. Furthermore, the multi-layered fuzzy rule engine integrates the membership degrees of behavior and textual signals, mapping the input information to a nonlinear risk space. Fuzzy inference introduces empirical rules, which can produce a gain effect when boundaries are ambiguous. For example, even slightly abnormal behavior but extremely sensitive text can trigger a high-risk warning, making the combined evaluation more discriminative than simple linear weighting, effectively improving the accuracy and robustness of electronic contract signing security identification.

[0075] The above describes the electronic contract risk identification method in the embodiments of the present invention. The following describes the electronic contract risk identification device in the embodiments of the present invention. Please refer to [link / reference]. Figure 4 One embodiment of the electronic contract risk identification device in this invention includes: The information acquisition module 401 is used to acquire the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; The feature extraction module 402 is used to extract the signing behavior sequence features of the signing behavior data and obtain the text content features of the electronic contract text; The risk identification module 403 is used to call the risk judgment model to identify risks in the electronic contract based on the signing behavior sequence features and the text content features.

[0076] The device provided in this embodiment of the invention can comprehensively consider the text of the electronic contract and the risk information in the signing process, thereby fully capturing the signing risks of the electronic contract and improving the sensitivity and accuracy of risk identification.

[0077] In another embodiment of this application, the feature extraction module 402 is further configured to: The multi-dimensional temporal action feature extraction model is invoked to extract the signature behavior sequence feature vector contained in the signature behavior data; The textual risk feature vector of the electronic contract text is extracted by calling a natural language processing model; The risk identification module 403 is specifically used for: The signature behavior sequence feature vector and the text risk feature vector are cross-fused to obtain fused vector information; The risk assessment model is invoked to identify risks in the electronic contract based on the fused vector information, and the risk identification results are obtained.

[0078] In another embodiment of this application, the risk assessment model is a multi-head attention structure; The step of cross-fusing the signature behavior sequence feature vector and the text risk feature vector to obtain the fused vector information includes: Calculate the similarity information between the signature behavior sequence feature vector and the text risk feature vector, respectively; The query dot product matrix is ​​obtained based on the similarity information, and an attention matrix is ​​constructed based on the query dot product matrix. Calculate the attention weights based on the attention matrix; Based on the multi-head attention structure and the attention weights, the signature behavior sequence feature vector and the text risk feature vector are weighted and cross-fused to obtain fused vector information.

[0079] In another embodiment of this application, the risk identification module 403 is specifically used to invoke the risk judgment model, perform risk assessment on the fused vector information based on the attention matrix, and output a risk assessment score.

[0080] In another embodiment of this application, the feature extraction module 402 is further configured to map the signing behavior sequence features and the text content features into graph nodes to construct a heterogeneous graph; Risk association information is obtained by performing information aggregation and multi-layer learning on information in the heterogeneous graphs based on graph convolution or graph attention networks. The risk identification module 403 is specifically used to call the risk judgment model to identify risks in the electronic contract based on the risk association information.

[0081] In another embodiment of this application, the electronic contract risk identification device further includes an optimization and update module, which is specifically used for: Obtain the manual review results for data deemed to contain risks, and annotate the data based on the review results; The risk assessment model is optimized and updated based on the labeled verification data.

[0082] The device provided in this embodiment of the invention can comprehensively consider the text of the electronic contract and the risk information in the signing process, thereby fully capturing the signing risks of the electronic contract and improving the sensitivity and accuracy of risk identification.

[0083] Based on the same inventive concept, this specification also provides an electronic device for identifying electronic contract risks. The electronic device for identifying electronic contract risks in this embodiment of the invention will be described in detail below from the perspective of hardware processing.

[0084] Figure 5 This is a schematic diagram of an electronic device provided as an embodiment of this specification. Refer to the following... Figure 5 To describe the electronic device 500 according to this embodiment of the invention. Figure 5 The electronic device 500 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0085] like Figure 5 As shown, the electronic device 500 is presented in the form of a general-purpose computing device. The components of the electronic device 500 may include, but are not limited to: at least one processing unit 510, at least one storage unit 520, a bus 530 connecting different system components (including storage unit 520 and processing unit 510), a display unit 540, etc.

[0086] The storage unit stores program code that can be executed by the processing unit 510, causing the processing unit 510 to perform the steps described in the processing method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 510 can perform, for example... Figure 1 , Figure 2 or Figure 3 The steps are shown.

[0087] The storage unit 520 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 5201 and / or a cache storage unit 5202, and may further include a read-only memory unit (ROM) 5203.

[0088] The storage unit 520 may also include a program / utility 5204 having a set (at least one) program module 5205, such program module 5205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0089] Bus 530 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0090] Electronic device 500 can also communicate with one or more external devices 100 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 500, and / or with any device that enables electronic device 500 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 550. Furthermore, electronic device 500 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 560. Network adapter 560 can communicate with other modules of electronic device 500 via bus 530. It should be understood that, although... Figure 5 As not shown, other hardware and / or software modules may be used in conjunction with the electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0091] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described in this invention can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this invention can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the method described above according to this invention. When the computer program is executed by a data processing device, it enables the computer-readable medium to implement the method described above, i.e.: as... Figure 1 , Figure 2 or Figure 3 The method shown.

[0092] Figure 6 This is a schematic diagram of a computer-readable medium provided for embodiments of this specification.

[0093] accomplish Figure 1 , Figure 2 or Figure 3 The computer program of the method shown can be stored on one or more computer-readable media. A computer-readable medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0094] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0095] In addition, the present invention also provides a computer program product, including a computer program / instruction that, when executed by a processor, implements the electronic contract risk identification method as described in any of the above embodiments.

[0096] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0097] In summary, the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that in practice, general-purpose data processing devices such as microprocessors or digital signal processors (DSPs) can be used to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0098] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the present invention is not inherently related to any specific computer, virtual device, or electronic device, and various general-purpose devices can also implement the present invention. The above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0099] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0100] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0101] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for identifying risks in electronic contracts, characterized in that, include: Obtain electronic contract texts and user signing behavior data generated when signing electronic contracts; Extract the signing behavior sequence features from the signing behavior data, and obtain the text content features of the electronic contract text; The risk assessment model is invoked to identify risks in the electronic contract based on the characteristics of the signing behavior sequence and the characteristics of the text content.

2. The electronic contract risk identification method according to claim 1, characterized in that, After extracting the signing behavior sequence features from the signing behavior data and obtaining the text content features of the electronic contract text, the method further includes: The multi-dimensional temporal action feature extraction model is invoked to extract the signature behavior sequence feature vector contained in the signature behavior data; The textual risk feature vector of the electronic contract text is extracted by calling a natural language processing model; The risk assessment model, based on the signature behavior sequence characteristics and the text content characteristics, identifies risks in the electronic contract, including: The signature behavior sequence feature vector and the text risk feature vector are cross-fused to obtain fused vector information; The risk assessment model is invoked to identify risks in the electronic contract based on the fused vector information, and the risk identification results are obtained.

3. The electronic contract risk identification method according to claim 2, characterized in that, The risk assessment model is a multi-head attention structure; The step of cross-fusing the signature behavior sequence feature vector and the text risk feature vector to obtain the fused vector information includes: Calculate the similarity information between the signature behavior sequence feature vector and the text risk feature vector, respectively; The query dot product matrix is ​​obtained based on the similarity information, and an attention matrix is ​​constructed based on the query dot product matrix. Calculate the attention weights based on the attention matrix; Based on the multi-head attention structure and the attention weights, the signature behavior sequence feature vector and the text risk feature vector are weighted and cross-fused to obtain fused vector information.

4. The electronic contract risk identification method according to claim 3, characterized in that, The risk assessment model is invoked to identify risks in the electronic contract based on the fused vector information, and the risk identification results include: The risk assessment model is invoked to perform risk assessment on the fused vector information based on the attention matrix, and a risk assessment score is output.

5. The electronic contract risk identification method according to claim 1, characterized in that, After extracting the signing behavior sequence features from the signing behavior data and obtaining the text content features of the electronic contract text, the method further includes: The signing behavior sequence features and the text content features are mapped to graph nodes to construct a heterogeneous graph; Risk association information is obtained by performing information aggregation and multi-layer learning on information in the heterogeneous graphs based on graph convolution or graph attention networks. The risk assessment model, based on the signature behavior sequence characteristics and the text content characteristics, identifies risks in the electronic contract, including: The risk assessment model is invoked to identify risks in the electronic contract based on the risk association information.

6. The method for identifying risks in electronic contracts according to any one of claims 1-5, characterized in that, Following the risk identification of the electronic contract, the method further includes: Obtain the manual review results for data deemed to contain risks, and annotate the data based on the review results; The risk assessment model is optimized and updated based on the labeled verification data.

7. An electronic contract risk identification device, characterized in that, The electronic contract risk identification device includes: The information acquisition module is used to acquire the electronic contract text and the signing behavior data generated by the user when signing the electronic contract; The feature extraction module is used to extract the signing behavior sequence features of the signing behavior data and obtain the text content features of the electronic contract text. The risk identification module is used to invoke the risk assessment model to identify risks in the electronic contract based on the signature behavior sequence characteristics and the text content characteristics.

8. An electronic contract risk identification device, characterized in that, The electronic contract risk identification device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic contract risk identification device to perform the steps of the electronic contract risk identification method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program / instructions thereon, characterized in that, When the program / instruction is executed by the processor, it implements the steps of the electronic contract risk identification method as described in any one of claims 1-6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, the steps of the electronic contract risk identification method as described in any one of claims 1-6 are implemented.

Images