Legal judgment prediction method based on deep extraction network and text structure analysis

Abstract

The present invention belongs to the technical field of legal judgment prediction, specifically a method for legal judgment prediction based on a deep extraction network and text structure analysis. The model employed includes an encoder, an attention layer, a deep extraction network, and a fact description separation module. The fact description and the crime definition are encoded to obtain fact description features and crime definition features. The fact description features and the crime definition features are subjected to contextual representation and an attention mechanism to obtain fact description features that incorporate legal knowledge. The fact description features that incorporate legal knowledge are subjected to a deep extraction network to obtain an abstract representation of the fact description, which is used to predict the crime. The predicted crime is encoded to obtain predicted crime features. The predicted crime features and the contextual representation of the fact description are subjected to a fact description separation module to obtain the legal basis and the sentence basis. The legal basis is subjected to another deep extraction network for feature extraction to obtain the legal basis semantic features, which are used to predict the legal provision. The predicted legal provision is encoded to obtain predicted legal provision features, which are used to predict the sentence. The deep extraction network extracts salient features and semantic information from the fact description, fully extracting the textual structural dependencies between the legal basis and the sentence basis in the fact description, thereby improving prediction accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of legal judgment prediction, and specifically is a legal judgment prediction method based on deep extraction network and text structure analysis. Technical Background

[0002] Legal judgment prediction aims to predict the judgment results of subtasks such as crime, legal provisions, and penalty period based on the description of case facts. The significant features in the fact description play a decisive role in judgment prediction, such as Figure 1 As shown in the figure, the gray background part is the salient features in the fact description. However, when the fact description passes through the deep network, these salient features will be weakened, and the semantic features between the contexts are also difficult to retain.

[0003] To leverage the key information in factual descriptions, some studies have used fine-grained key event information extracted from factual descriptions to match event types defined in legal provisions. This allows accurate retrieval of legal provisions applicable to the case and the use of judgments in legal provisions to infer sentencing terms. However, this approach ignores the semantic information between the contextual elements in the factual descriptions. Other studies have used deep neural networks to extract key information from factual descriptions for legal judgment prediction. However, the feature information in factual descriptions is easily weakened when passing through deep neural networks, thus affecting the prediction results.

[0004] In actual judicial scenarios, judges usually refer to various parts of the factual description (legal basis and sentence basis) to decide the verdict. Figure 1 As shown in the figure, the solid and dashed lines represent the legal basis and sentence length, respectively, for the judge's decision. Some studies have used generative models to derive the basis for a verdict from factual descriptions and then make predictions based on these generated basis. Other studies have used the results of intermediate subtasks to separate factual descriptions into different contexts and then use these to make predictions for other subtasks.

[0005] The problem with the above method is that it fails to account for the inherent structure of factual descriptions. For example, the legal basis often appears before the sentence basis. This inherent structure not only reflects the rigor and standardization of legal document writing, but also contains important predictive information. Ignoring this structure can lead to information omissions and reduced prediction accuracy. Therefore, this paper proposes a legal judgment prediction method based on deep extraction networks and text structure analysis. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the technical problem that the present invention intends to solve is to propose a legal judgment prediction method based on deep extraction network and text structure analysis.

[0007] The present invention solves the technical problem by adopting the following technical solutions:

[0008] A method for predicting legal decisions based on deep extraction networks and text structure analysis, characterized in that the method comprises the following steps:

[0009] Step 1: Build a legal judgment prediction model, including an encoder, an attention layer, a deep extraction network, and a fact description separation module;

[0010] The fact description and crime definition are input into the encoder for encoding, respectively, to obtain fact description features and crime definition features. The fact description features and crime definition features are subjected to maximum pooling operations to obtain contextual representations of the fact description and crime definition. The contextual representations of the fact description and crime definition are passed through the attention layer to obtain fact description features that incorporate legal knowledge.

[0011] The fact description features that incorporate legal knowledge are input into a deep extraction network to extract the salient features and semantic information in the fact description to obtain an abstract representation of the fact description. The abstract representation of the fact description is passed through the softmax function to obtain the probability distribution of the predicted crime. The probability distribution of the predicted crime is passed through the argmax function to obtain the predicted crime.

[0012] The predicted crime is encoded using an encoder to obtain the predicted crime features. The predicted crime features and the contextual representation of the factual description are passed through the fact description separation module to predict the legal basis and the sentence basis, thereby obtaining the legal basis and the sentence basis. The legal basis is subjected to feature extraction through another deep extraction network to obtain the semantic features of the legal basis. The semantic features of the legal basis are passed through the softmax function to obtain the probability distribution of the predicted legal provision. The probability distribution of the predicted legal provision is passed through the argmax function to obtain the predicted legal provision.

[0013] The predicted legal article is encoded using an encoder to obtain the predicted legal article features; the predicted legal article features are concatenated with the sentence based on semantic features, and then the sentence is predicted using a softmax function to obtain the predicted sentence;

[0014] Step 2: Train the legal judgment prediction model and use the trained legal judgment prediction model for legal judgment prediction.

[0015] Furthermore, the depth extraction network includes multiple serially connected depth extraction modules, and the output features of each depth extraction module are spliced ​​to obtain the output features of the depth extraction network; the input features of the depth extraction module are sequentially upsampled, Inception module A and Inception module B, and the output features of Inception module B are residually connected with the upsampled features, and then undergo a K-top pooling operation and a fully connected layer to obtain the output features of the depth extraction module;

[0016] The feature extraction process of Inception module A is expressed as:

[0017] R A =Concat(C 1×1 (F),C 1×3 (C 1×1 (F)),C 1×7 (C 1×1 (F)),C avg (C 1×1 (F))) (6)

[0018] Where R A is the output feature of Inception module A, F is the input feature of Inception module A, Concat represents the concatenation operation, C 1×1 、C 1×3 and C 1×7 They are 1×1, 1×3 and 1×7 convolution operations, C avg is the average pooling operation;

[0019] The feature extraction process of Inception module B is expressed as:

[0020] R B =Concat(C 1×1 (R A ),C 1×2 (C 1×1 (R A )),C 1×4 (C 1×1 (R A )),C avg (C 1×1 (R A )))(7)

[0021] Where R B is the output feature of Inception module B, C 1×2 、C 1×4 Represent 1×2 and 1×4 convolution operations respectively.

[0022] Furthermore, the fact description separation module includes a Transformer layer and a CRF layer. The Transformer layer uses the predicted crime features as the key vector and value vector, and the contextual representation of the fact description as the query vector, and generates a contextual representation from the crime to the fact through the attention mechanism; the contextual representation from the crime to the fact is feature mapped through the softmax function to obtain the probability distribution of the legal basis and the basis of the sentence; the probability distribution of the legal basis and the basis of the sentence is input into the CRF layer to generate the optimal label sequence of the legal basis and the basis of the sentence; the legal basis and the basis of the sentence are separated from the optimal label sequence of the legal basis and the basis of the sentence.

[0023] Furthermore, the generation process of the fact description feature incorporating legal knowledge is expressed as follows:

[0024]

[0025] Where H is the fact description feature integrated with legal knowledge, is the contextual representation of the fact description, c is the contextual representation of the crime definition, W is the weight coefficient matrix between the fact description and the crime definition, W u is a learnable parameter matrix.

[0026] Furthermore, during the model training process, the training loss is calculated by the following formula:

[0027]

[0028] Where λ j represents the weight factor of subtask j, j = 1, 2, 3 correspond to the three subtasks of law prediction, crime prediction and sentence prediction respectively, K j represents the number of samples of subtask j, y j,k 、 They represent the true label of the kth sample in subtask j respectively.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. A deep extraction network (MDEN) was designed to extract salient features and semantic information from factual descriptions and legal basis, which helps improve the accuracy and reliability of predictions. Considering the structural nature of factual descriptions, namely the textual dependencies between the legal basis and the sentence basis in factual descriptions, this structurality was analyzed through factual description separation. Conditional random fields (CRFs) were used to more accurately capture the logical order and interdependencies between elements such as the legal basis and the sentence basis, improving the accuracy of factual description separation and, in turn, the accuracy of legal judgment prediction. The combination of deep extraction networks and factual description separation not only improves the prediction accuracy of legal judgments but also enriches deep learning technology's understanding of the characteristics of legal documents.

[0031] 2. In the deep extraction module, upsampling is used to enhance input features, abstract features are learned through convolutional networks, and noisy features are filtered using K-top pooling operations. Through in-depth analysis and extraction of salient features and semantic information, the salient features and semantic information in factual descriptions and legal basis are retained when passing through the deep extraction network, thereby improving prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is an example of a description of the facts of the case;

[0033] Figure 2 is a structural diagram of the legal decision prediction model of the present invention;

[0034] Figure 3 is a structural diagram of the depth extraction network of the present invention;

[0035] Figure 4 is a structural diagram of the depth extraction module of the present invention;

[0036] Figure 5 It is a schematic diagram of the relative positions of the legal basis and the sentencing basis in the factual description. DETAILED DESCRIPTION

[0037] Specific embodiments are given below in conjunction with the accompanying drawings. The specific embodiments are only used to introduce the technical solutions of the present invention in detail and are not intended to limit the scope of protection of the present application.

[0038] The present invention provides a legal judgment prediction method based on deep extraction network and text structure analysis (hereinafter referred to as method, see Figures 1 to 5 ), including the following steps:

[0039] Step 1: Build a legal judgment prediction model; the legal judgment prediction model includes an encoder, an attention layer, a deep extraction network, and a fact description separation module;

[0040] The fact description and crime definition are input into the encoder respectively. The encoder uses the BERT model to encode the fact description and crime definition to obtain the fact description features and crime definition features. The encoding process of the fact description is expressed as:

[0041] [h1,...,h n ]=BERT[x1,...,x n ] (1)

[0042] Where, [h1,...,h n ] is a sequence of fact description features, BERT is a BERT model, [x1,...,x n ] is a text sequence describing facts, and n is the length of the fact description;

[0043] The fact description features and the crime definition features are subjected to the maximum pooling operation respectively to obtain the context representation of the fact description and the crime definition; among them, the context representation of the fact description for:

[0044]

[0045] In the formula, maxpooling is the maximum pooling operation;

[0046] The contextual representations of the fact description and the crime definition are input into the attention layer. The contextual representation of the crime definition is used to query the contextual representation of the fact description in order to mine the semantic information most relevant to the crime definition in the fact description. The legal knowledge is then integrated into the fact description through the attention mechanism to obtain the fact description features that incorporate legal knowledge. The process is expressed as follows:

[0047]

[0048] Where H is the fact description feature integrated with legal knowledge, c is the context representation of the crime definition, W is the weight coefficient matrix between the fact description and the crime definition, and W u is a learnable parameter matrix;

[0049] The factual description features, incorporating legal knowledge, are input into a deep extraction network (MDEN). This extracts salient features and semantic information from the factual descriptions, resulting in an abstract representation of the factual descriptions. This abstract representation encodes multi-granular semantic information and semantic dependencies, which are then used for charge prediction. The deep extraction network comprises multiple deep extraction modules (DENs), five in this embodiment. The output features of the previous deep extraction module serve as the input features for the next deep extraction module. The output features of each deep extraction module are concatenated to form the output features of the deep extraction network.

[0050] In the depth extraction module, the input features are sequentially upsampled, passed through Inception module A, and passed through Inception module B. The output features of Inception module B are residually connected with the upsampled features, and then pass through K-top pooling and a fully connected layer to obtain the output features of the depth extraction module.

[0051] Upsampling is an enhancement of input features to facilitate the subsequent extraction of richer feature information. Upsampling only copies the input features, so it does not increase the number of model parameters. The upsampling operation is expressed by the following formula:

[0052] [H i1 ,...,H iv ]=Upsample(H i )(5)

[0053] Where H i is the input feature of the i-th deep feature extraction module, [H i1 ,-,H iv ] is the up-sampled feature, H i1 =H iv =H i , v represents the predefined parameters of upsampling, and Upsample represents the upsampling operation.

[0054] Inception modules A and B are designed to extract features of different granularities from input data using convolutions of different scales. The input features of Inception module A are extracted through four parallel 1×1 convolutions. The output features of the three 1×1 convolutions are then average pooled, 1×3 convolutions, and 1×7 convolutions before being concatenated with the output features of the last 1×1 convolution to provide a richer feature representation for the next module. The output features of Inception module A are obtained. The feature extraction process of Inception module A is expressed as follows:

[0055] R A =Concat(C 1×1 (F),C 1×3 (C 1×1 (F)),C 1×7 (C 1×1 (F)),C avg (C 1×1 (F)))(6)

[0056] Where R A is the output feature of Inception module A, Concat represents the concatenation operation, C 1×1 、C 1×3 and C 1×7They are 1×1, 1×3 and 1×7 convolution operations respectively; C avg is an average pooling operation, which is used to reduce the feature dimension and retain important information; F = [H i1 ,...,H iv ] is the input feature of the Inception module A, that is, the feature obtained by upsampling.

[0057] Similarly, the input features of Inception module B are extracted through four parallel 1×1 convolutions. The output features of the three 1×1 convolutions are respectively subjected to average pooling, 1×2 convolution, and 1×4 convolution, and then concatenated with the output features of the last 1×1 convolution to obtain the output features of Inception module B. The feature extraction process of Inception module B is expressed as:

[0058] R B =Concat(C 1×1 (R A ),C 1×2 (C 1×1 (R A )),C 1×4 (C 1×1 (R A )),C avg (C 1×1 (R A )))(7)

[0059] Where R B is the output feature of Inception module B, C 1×2 、C 1×4 Represent 1×2 and 1×4 convolution operations respectively;

[0060] In order to further enhance the feature transfer capability of the model and prevent the gradient vanishing problem during training, the output features of the Inception module B are connected to the features obtained by upsampling through residual connections, which enables rapid information flow between features at different levels, thereby helping to improve training results and accelerate the convergence process. The residual connection is expressed as:

[0061] R=F+R B (8)

[0062] Where R is the feature obtained by residual connection.

[0063] The features obtained by residual connection are subjected to K-top pooling operation and full connection layer to obtain the output features of the depth extraction module. The process is expressed as follows:

[0064] H i '=σ(K-top(R))(9)

[0065] Where H i ' is the output feature of the i-th depth extraction module, σ represents the fully connected layer, and K-top represents the K-top pooling operation.

[0066] The abstract representation of the fact description is passed through the softmax function to obtain the probability distribution of the predicted crime; the probability distribution of the predicted crime is passed through the argmax function to obtain the predicted crime. The process is expressed as follows:

[0067]

[0068] Where, is the probability distribution of the predicted crime, softmax is the softmax function, Q is the abstract representation of the fact description, and W c 、b c is the learnable parameter matrix and bias, argmax is the argmax function, It is to predict the crime;

[0069] Using encoders to predict crimes Encode and obtain the predicted crime characteristics Taking into account the text structure of the fact description, the crime characteristics will be predicted and contextual representation of fact descriptions Input into the fact description separation module to predict the legal basis and sentence basis, and obtain the legal basis H a and sentence based on H t , used for the prediction of laws and crimes.

[0070] The fact description separation module includes a Transformer layer and a CRF layer. The Transformer layer uses the predicted crime features as the key vector and value vector, and the contextual representation of the fact description as the query vector. It uses the inherent connection between the crime definition and the fact description to generate a contextual representation from the crime to the fact. Its operation is expressed as follows:

[0071]

[0072] Where D is the context representation from the charge to the fact, Attention represents the attention mechanism, K and V are the key vector and value vector, Q is the query vector, and d k is the dimension of the key vector and value vector;

[0073] Use the softmax function to map the contextual representation from the crime to the fact into the feature space, and predict the probability distribution of the legal basis and the sentence basis;

[0074] P=softmax(W s D+b s )(13)

[0075] Where P is the probability distribution of legal basis and sentence basis, W s 、b s is the learnable parameter matrix and bias;

[0076] The probability distribution P of the legal basis and the sentencing basis is input into the CRF layer to generate the optimal label sequence of the legal basis and the sentencing basis. The CRF layer can learn the transition relationship between labels. By considering the transition probability between labels and the conditional probability of each label, it can capture the dependency relationship in the label sequence. For example, the legal basis should be after the sentencing basis. The generation process of the optimal label sequence is expressed as:

[0077]

[0078] Where, The optimal label sequence representing the legal basis and the sentence basis; is the element of the state transition matrix, representing the transition from the previous label y i-1 Go to the current label y i The state transition matrix is ​​the learnable parameter matrix of the CRF layer, which captures the transition probability between adjacent labels in the label sequence; Indicates the label y at position i i The conditional probability of

[0079] The optimal label sequence based on legal provisions and sentence terms Separation of legal provisions according to H a and sentence based on H t ; Legal basis H a After feature extraction by another deep extraction network (MDEN), the legal clause semantic feature H encoding multi-granularity semantic features and semantic dependency relationships is obtained. a '; The legal provisions are based on semantic features H a 'The probability distribution of the predicted legal article is obtained through the softmax function, and the probability distribution of the predicted legal article is obtained through the argmax function. The process is expressed as:

[0080]

[0081] Where, It is the prediction law, W a 、b a is the learnable parameter matrix and bias, It is a predictive law;

[0082] Use the encoder to encode the predicted legal provisions to obtain the predicted legal provisions features The predicted legal feature and the sentence semantic feature H t 'Splicing, and then using the softmax function to predict the sentence, to get the predicted sentence; the process is expressed as:

[0083]

[0084] Where, is the predicted sentence, W t 、b t is a learnable parameter matrix and bias, [;] represents the concatenation operation;

[0085] Step 2: Train the legal judgment prediction model, calculate the training loss according to the following formula, and use the trained legal judgment prediction model for legal judgment prediction;

[0086]

[0087] Where λ j represents the weight factor of subtask j, j = 1, 2, 3 correspond to the three subtasks of law prediction, crime prediction and sentence prediction respectively, K j represents the number of samples of subtask j, y j,k 、 They represent the true label of the kth sample in subtask j respectively.

[0088] Example

[0089] This example was tested on the CAIL-big and CAIL-small datasets. First, the datasets were preprocessed to filter out infrequent crimes and legal provisions, retaining only legal provisions with a frequency greater than 100. The sentences were then divided into non-overlapping intervals to filter out multi-label samples.

[0090] Through in-depth analysis of the CAIL-small dataset, we show the structure of legal fact descriptions. Due to the lack of corresponding labeled data, we randomly selected 50 cases from CAIL-small and labeled them with the legal basis and the sentence basis. Figure 5 The figure shows the relative positions of the two types of evidence within the factual description. The horizontal axis (0%) indicates the beginning of the factual description, and the vertical axis (100%) indicates the end. The vertical axis represents the number of factual descriptions. The blue and yellow colors represent the legal basis and sentencing basis, respectively, within the factual description. For example, the blue area between 0% and 70% of a given case indicates that the legal basis is located between 0% and 70% of the factual description. The figure shows that the relative positions of the legal basis and sentencing basis within the factual description are stable. This result suggests that the structural nature of the factual description can be considered to separate the factual descriptions.

[0091] For the encoder (BERT model), the maximum length of the fact description is set to 512. To optimize the parameters, the Adam optimizer is used, the learning rate is set to 0.001, and the dropout rate is set to 0.5. The legal judgment prediction model is implemented using PyTorch and trained on a server using 2×H800 GPUs. Each model is trained for 16 epochs with a batch size of 128.

[0092] In order to verify the effectiveness of the legal judgment prediction model of the present invention, Word2Vec+SVM was selected. [1-2] 、FAL [3] TOPJUDGE [4] 、LADAN [5] NeurJudge [6] , CTM [7] and EMP [8] We use it as a baseline model for comparison and evaluate it using four metrics: Accuracy (Acc), Macro Precision (MP), Macro Recall (MR), and Macro F1 score. Tables 1 and 2 show the experimental results on the CAIL-small and CAIL-big datasets.

[0093] Table 1 Comparison of experimental results between the proposed model and the baseline model on the CAIL-small dataset

[0094]

[0095]

[0096] Table 2 Comparison of experimental results between the proposed model and the baseline model on the CAIL-big dataset

[0097]

[0098] The above results show that all models outperformed the CAIL-big dataset on the CAIL-small dataset. This is because the CAIL-big dataset is larger than the CAIL-small dataset. Larger datasets typically contain more diverse samples, which helps the models learn more extensive and complex feature representations, enabling them to better generalize to unseen data. Compared to the baseline model, the proposed method improved all metrics on all three tasks to a certain extent, maintaining relatively stable precision and recall. In particular, the F1 score was the highest across all three tasks, demonstrating the model's robustness and generalization ability. Although none of the models achieved very high accuracy on the sentence prediction task, the proposed model still achieved the best performance (Acc of 40.27% and 61.53%, respectively), indicating that sentence prediction can be a more complex and challenging task. The proposed method analyzes the structure of factual descriptions through fact description separation, identifying and understanding the logical relationships and organizational structure of each component of the factual description. It effectively captures the legal basis and sentence basis, improving the accuracy and interpretability of sentence prediction. The deep extraction network can extract key information and semantic dependencies from the text, further improving the accuracy of prediction.

[0099] To verify the effectiveness of the legal knowledge fusion, deep extraction network, and fact description separation of the present invention, ablation experiments were conducted on these three parts on the CAIL-small dataset. The results are shown in Table 3.

[0100] Table 3 Ablation experiment results on CAIL-small dataset

[0101]

[0102]

[0103] In Table 3, "-Att" indicates that the factual description is not integrated with the crime definition, and the factual description is used directly for prediction; "-MDEN" indicates that the deep extraction network is not used to extract salient features and semantic information; and "-Separation" indicates that the factual description separation is not performed. Ablation experiment results show that after removing "Att" or "MDEN", ​​all indicators in the three tasks show varying degrees of decline, verifying the effectiveness of integrating the factual description with the crime definition and extracting salient features and semantic information. After removing "Separation", all indicators in the two subtasks of legal provision prediction and sentence prediction show varying degrees of decline, confirming the effectiveness of factual description separation for legal provision and sentence prediction.

[0104] The sources of the above models are as follows:

[0105] [1]Mikolov T,Sutskever I,Chen K,et al.Distributed representations ofwords and phrases and their compositionality[J].Advances in neuralinformation processing systems,2013,26.

[0106] [2]Suykens J A K,Vandewalle J.Least squares support vector machineclassifiers[J].Neural processing letters,1999,9:293-300.

[0107] [3]Luo B,Feng Y,Xu J,et al.Learning to predict charges for criminalcases with legal basis[J].

[0108] arxiv preprint arxiv:1707.09168,2017.

[0109] [4]Zhong H,Guo Z,Tu C,et al.Legal judgment prediction via topologicallearning[C] / / Proceedings of the 2018conference on empirical methods innatural language processing.2018:3540-3549.[5]Xu N,Wang P,Chen L,etal.Distinguish confusing law articles for legal judgment prediction[J].

[0110] arxiv preprint arxiv:2004.02557,2020.

[0111] [6]Yue L,Liu Q,Jin B,et al.Neurjudge:A circumstance-aware neural framework for legal judgment prediction[C] / / Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval.2021:973-982.

[0112] [7]CHEN H,CAID,DAI W,et al.Charge-Based Prison Term Prediction with Deep Gating Network[C / OL] / / Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing(EMNLP-IJCNLP),2019.

[0113] [8]Feng Y,Li C,Ng V.Legal judgment prediction via event extraction with constraints[C] / / Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics(Volume 1:Long Papers).2022:648-664.

[0114] The parts not described in this invention are applicable to the prior art.

Claims

1. A legal judgment prediction method based on deep extraction network and text structure analysis, characterized in that: The method comprises the following steps: Step 1: Build a legal judgment prediction model, including an encoder, an attention layer, a deep extraction network, and a fact description separation module; The fact description and crime definition are input into the encoder for encoding, respectively, to obtain fact description features and crime definition features. The fact description features and crime definition features are subjected to maximum pooling operations to obtain contextual representations of the fact description and crime definition. The contextual representations of the fact description and crime definition are passed through the attention layer to obtain fact description features that incorporate legal knowledge. The fact description features that incorporate legal knowledge are input into a deep extraction network to extract the salient features and semantic information in the fact description to obtain an abstract representation of the fact description. The abstract representation of the fact description is passed through the softmax function to obtain the probability distribution of the predicted crime. The probability distribution of the predicted crime is passed through the argmax function to obtain the predicted crime. The predicted crime is encoded using an encoder to obtain the predicted crime features. The predicted crime features and the contextual representation of the factual description are passed through the fact description separation module to predict the legal basis and the sentence basis, thereby obtaining the legal basis and the sentence basis. The legal basis is subjected to feature extraction through another deep extraction network to obtain the semantic features of the legal basis. The semantic features of the legal basis are passed through the softmax function to obtain the probability distribution of the predicted legal provision. The probability distribution of the predicted legal provision is passed through the argmax function to obtain the predicted legal provision. The predicted legal article is encoded using an encoder to obtain the predicted legal article features; the predicted legal article features are concatenated with the sentence based on semantic features, and then the sentence is predicted using a softmax function to obtain the predicted sentence; The depth extraction network includes multiple serially connected depth extraction modules. The output features of each depth extraction module are spliced ​​to obtain the output features of the depth extraction network. The input features of the depth extraction module are sequentially upsampled, Inception module A and Inception module B. The output features of Inception module B are residually connected with the upsampled features, and then undergo a K-top pooling operation and a fully connected layer to obtain the output features of the depth extraction module. The feature extraction process of Inception module A is expressed as: R A =Concat(C 1×1 (F),C 1×3 (C 1×1 (F)),C 1×7 (C 1×1 (F)),C avg (C 1×1 (F)))(6) Where R A is the output feature of Inception module A, F is the input feature of Inception module A, Concat represents the concatenation operation, C 1×1 、C 1×3 and C 1×7 They are 1×1, 1×3 and 1×7 convolution operations, C avg is the average pooling operation; The feature extraction process of Inception module B is expressed as: R B =Concat(C 1×1 (R A ),C 1×2 (C 1×1 (R A )),C 1×4 (C 1×1 (R A )),C avg (C 1×1 (R A )))(7) Where R B is the output feature of Inception module B, C 1×2 、C 1×4 Represent 1×2 and 1×4 convolution operations respectively; Step 2: Train the legal judgment prediction model and use the trained legal judgment prediction model for legal judgment prediction.

2. The legal judgment prediction method based on deep extraction network and text structure analysis according to claim 1 is characterized in that: The fact description separation module includes a Transformer layer and a CRF layer. The Transformer layer uses the predicted crime features as the key vector and value vector, and the contextual representation of the fact description as the query vector, and generates a contextual representation from the crime to the fact through the attention mechanism; the contextual representation from the crime to the fact is mapped through the softmax function to obtain the probability distribution of the legal basis and the sentence basis; the probability distribution of the legal basis and the sentence basis is input into the CRF layer to generate the optimal label sequence of the legal basis and the sentence basis; the legal basis and the sentence basis are separated from the optimal label sequence of the legal basis and the sentence basis.

3. The legal judgment prediction method based on deep extraction network and text structure analysis according to claim 2 is characterized in that: The generation process of the fact description features incorporating legal knowledge is expressed as follows: Where H is the fact description feature integrated with legal knowledge, is the contextual representation of the fact description, c is the contextual representation of the crime definition, W is the weight coefficient matrix between the fact description and the crime definition, W u is a learnable parameter matrix.

4. The legal judgment prediction method based on deep extraction network and text structure analysis according to claim 1 is characterized in that: During model training, the training loss is calculated as follows: Where λ j represents the weight factor of subtask j, j = 1, 2, 3 correspond to the three subtasks of law prediction, crime prediction and sentence prediction respectively, K j represents the number of samples of subtask j, y j,k 、 They represent the true label of the kth sample in subtask j respectively.

Images