CN113487453B - Legal judgment prediction method and system based on crime elements - Google Patents
Legal judgment prediction method and system based on crime elements Download PDFInfo
- Publication number
- CN113487453B CN113487453B CN202110632603.3A CN202110632603A CN113487453B CN 113487453 B CN113487453 B CN 113487453B CN 202110632603 A CN202110632603 A CN 202110632603A CN 113487453 B CN113487453 B CN 113487453B
- Authority
- CN
- China
- Prior art keywords
- sentence
- criminal
- representation
- lstm
- representations
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000013598 vector Substances 0.000 claims abstract description 27
- 230000002787 reinforcement Effects 0.000 claims abstract description 12
- 230000006399 behavior Effects 0.000 claims abstract description 7
- 239000003795 chemical substances by application Substances 0.000 claims description 31
- 230000009471 action Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 12
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 4
- 238000002372 labelling Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims 2
- 238000004590 computer program Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/18—Legal services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- Biomedical Technology (AREA)
- Technology Law (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a law judgment prediction method and a law judgment prediction system based on criminals, which utilize layering Bi-LSTM to carry out fact description coding and generate sentence representation of a context; extracting elements based on reinforcement learning, and selecting sentences containing crime elements; generating a corresponding discriminating criminal element representation by fusing the contextual representations of the selected sentences; and splicing criminal element representations for distinguishing criminals, criminal targets, criminal intentions and criminal behaviors, and performing multitask judgment and prediction on the formed spliced vectors to obtain a prediction judgment result. The invention can effectively distinguish confusable crime facts and legal clauses, and improves the accuracy of legal judgment and prediction tasks.
Description
Technical Field
The invention belongs to the technical field of legal judgment prediction, and particularly relates to a legal judgment prediction method and system based on crime elements.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Legal decision prediction aims at predicting decision results (e.g. legal terms, applicable criminal names, criminal periods, etc.) based on case facts. Early studies on legal decision prediction focused on using statistical solutions, and recent studies on legal decision prediction mostly treated this task as a multi-classification problem. Various text classification methods have been applied to address legal decision prediction tasks. However, this task still presents two challenges: (1) Many of the fact descriptions in legal decision predictions are similar, (2) a large number of criminal names and their applicable legal terms are semantically similar. Due to the two challenges, the existing legal decision prediction method is easy to misdecide legal cases.
Disclosure of Invention
In order to solve the problems, the invention provides a law judgment prediction method and a system based on crime elements.
According to some embodiments, the present invention employs the following technical solutions:
a law judgment prediction method based on crime elements comprises the following steps:
carrying out fact description coding by using the layering Bi-LSTM, and generating sentence representation of the context;
extracting elements based on reinforcement learning, and selecting sentences containing crime elements;
generating a corresponding discriminating criminal element representation by fusing the contextual representations of the selected sentences;
and splicing criminal element representations for distinguishing criminals, criminal targets, criminal intentions and criminal behaviors, and performing multitask judgment and prediction on the formed spliced vectors to obtain a prediction judgment result.
As an alternative embodiment, the specific process of fact description encoding using hierarchical Bi-LSTM includes: with four hierarchical Bi-LSTMs, each having two levels of Bi-LSTMs, the word level Bi-LSTMs output contextual word representations, the basic sentence representations are computed by the attention mechanism, and the contextual sentence representations are output by the sentence level B i-LSTMs.
As a further defined embodiment, each fact description has a plurality of sentences, each sentence comprising a plurality of words, for each input sentence in the fact description, a sequence of contextual word representations is output using word-level Bi-LSTM, a word-level attention vector is calculated based on the sequence of contextual word representations and the word-level context vector, a basic representation of a sentence is obtained, a basic sentence representation sequence is calculated based on the word-level attention mechanism, and sentence-level Bi-LSTM is used to calculate the sequence of contextual sentence representations.
As an alternative embodiment, the specific process of extracting the elements based on reinforcement learning includes: four agents are used to select sentences containing criminals of different types, including criminals, criminal objectives, criminal intent, and criminal activity, respectively, each agent employing a random strategy to sample the actions in each state.
As a further defined embodiment, the accuracy of crime extraction is measured using different delay rewards, a first delay reward is provided for a kth agent by determining the difference between the kth crime extracted by the agent and the true kth crime, a second delay reward is provided for all agents by calculating a prediction error relative to the true labeling legal terms, and the final reward for each agent is calculated using the sum of the two delay rewards.
As an alternative embodiment, the specific process of contextual representation by fusing selected sentences includes: the context representations of sentences containing crimes are fused to generate a kth discriminative crime representation, a sentence-level attention weight vector is calculated based on the action sequence, the context sentence representations and the sentence-level context vector, and a crime representation is calculated based on the attention weight vector.
As an alternative embodiment, the specific process of performing the multitasking decision prediction on the formed spliced vector includes: the representations of the four crime elements are linked together as representations of the fact descriptions, and a fact representation is generated for the respective legal decision prediction subtasks based on the representations of the fact descriptions.
A crime element-based legal decision prediction system, comprising:
a fact description encoder configured to encode the fact description using the hierarchical Bi-LSTM, generating a sentence representation of the context;
an element extractor configured to extract elements by reinforcement learning, and to pick out sentences containing crime elements;
a crime element discriminator configured to generate a corresponding discriminated crime element representation by fusing the contextual representations of the selected sentences;
the multi-task judgment predictor is configured to splice criminal element representations for distinguishing criminals, criminal targets, criminal intentions and criminal behaviors, and to perform multi-task judgment prediction on formed spliced vectors to obtain a prediction judgment result.
A computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to perform the steps of the above method.
A terminal device comprising a processor and a computer readable storage medium, the processor configured to implement instructions; the computer readable storage medium is for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of the above-described method.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a legal judgment prediction system based on crime elements for the first time, and by extracting the crime element information with distinguishing property, the system can effectively distinguish confusing crime facts and legal terms, thereby improving the accuracy index of legal judgment prediction tasks.
The present invention innovatively proposes to use reinforcement learning based crime element extractors to discover crime elements (including criminals and crime targets) to distinguish confusing fact descriptions and to use element discriminators to distinguish legal terms with similar TF-I DF representations to facilitate prediction of legal terms by an overall model.
In the application level, a large number of experiments performed on the reference data set prove the effectiveness of the proposed legal decision prediction system. Crime element-based legal decision prediction systems may provide assistance to ordinary people who are unfamiliar with legal terms and complex legal consultation procedures. In addition, the legal judgment prediction result can also be used as a reference of professionals (such as lawyers and judges), so that the working efficiency of legal staff is improved.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a schematic diagram of a system according to the present invention.
The specific embodiment is as follows:
the invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
The invention provides a legal judgment prediction system based on crime elements, which mainly aims at extracting four specific crime elements in crime facts and distinguishing confusing crime facts and legal terms according to an information model of the crime elements.
The legal judgment prediction system based on crime elements provided by the invention comprises four parts. As shown in fig. 1, includes (1) a fact description encoder, (2) a reinforcement learning-based element extractor, (3) a crime element discriminator, and (4) a multitasking decision predictor.
The present invention utilizes a hierarchical Bi-LSTM as a fact description encoder to generate sentence representations of a context. Next, in order to distinguish confusing fact descriptions, the present invention utilizes an element extractor based on reinforcement learning to accurately pick sentences containing crime elements. Then, to enhance the prediction of legal terms, the present invention employs a crime element discriminator that generates a kth discriminated crime element representation by fusing the contextual representations of the selected sentencesWhere k=1, 2,3,4. After that, four discriminant criminals are expressed +.>Is input to a multi-task decision predictor to accurately predict a decision result.
Various portions of the crime element-based legal decision prediction system are described in detail below.
1. Fact description encoder
To model the contextual sentence representations in the fact description, the present invention employs four hierarchical Bi-LSTM. Each hierarchical Bi-LSTM has two levels of Bi-LSTM, a word level Bi-LSTM outputs a contextual word representation, then a basic sentence representation is computed by the attention mechanism, and finally a sentence level Bi-LSTM outputs a contextual sentence representation. Each fact describes a plurality of sentences, each sentence containing a plurality of words. Specifically, for each input sentence in the fact description fWherein n is i Referring to the number of words in the i-th sentence, the word level Bi-LSTM outputs a sequence of contextual word representations, namely:
wherein the method comprises the steps ofw i,j Is->Is represented by the j-th basic word and is initialized by the Skip-Gram model.K=1, 2,3,4, representing a learnable parameter of the kth word level Bi-LSTM. Representing sequences and word-level context vectors based on context words +.>Calculating a word level attention vector +.>
Wherein the method comprises the steps ofIs a trainable weight matrix of word level attention mechanisms. Then obtain a basic representation of a sentenceNamely:
based on word level attention mechanism, basic sentence representation sequence is calculatedWherein n is f Representing the number of sentences in the fact description f. The sentence-level Bi-LSTM is then used to calculate a context sentence representation sequence, namely:
wherein the method comprises the steps of K=1, 2,3,4, representing a learnable parameter of the kth sentence level Bi-LSTM.
2. Element extractor based on reinforcement learning
In order to distinguish confusing fact descriptions, the invention adopts four intelligent agents to respectively select sentences containing different types of criminal elements. Each agent adopts a random strategy pi k To sample the actions in each state. Based on the sampled actions, the kth agent determines whether the sentence contains the kth crime element, k=1, 2,3,4. Wherein the actions, status, policies and rewards of the agent are as follows:
agent action
ActionSelected from the agent action space { positive=1, negative=0 } where positive indicates that the sentence contains the kth crime element and negative indicates that the sentence does not contain the kth crime element.
State of agent
The present invention uses a recurrent neural network to encode agent state representations:
wherein the method comprises the steps ofT th sentence S t In the context of t.epsilon.1, n f ]，/>K=1, 2,3,4, representing the learnable parameters of the kth GRU. Note that at the beginning (t=1), the state +.>Is a trainable vector.
2.3 agent policy
A random strategy for extracting kth criminals, which specifies a probability distribution of agent actions:
wherein the method comprises the steps ofA learnable parameter representing a kth crime element extraction policy.
2.4 agent rewards
Since there is no label as to which sentences contain criminals, the agent, after scanning all sentences in the fact description, devised two delay rewards to measure the correctness of criminal extraction. Specifically, the crime element discriminator provides a first type of delay rewards for the kth agent by discriminating the kth crime element extracted by the agent from the true kth crime element
Wherein the method comprises the steps ofTrue annotation vector representing kth crime element,/->Refers to the discrimination result of the kth crime element, k=1, 2,3,4./>And->Refers to the number of categories of the kth crime element. The correctness of four law-related criminals affects the outcome of the legal provision prediction task. Thus, the legal provision predictor provides a second type of delay incentive r for all agents by calculating a prediction error relative to the true labeling legal provision p ：
Wherein the method comprises the steps ofIs a ground truth vector referring to law clause prediction, < ->Predictive outcome representing legal terms, mε [1, |Y 1 |]，|Y 1 I is the number of legal terms. The final prize for the kth agent is calculated from the sum of the two delayed prizes, namely:
3. crime element discriminator
To enhance the prediction of legal terms, the present invention fuses the contextual representation of sentences containing crime elements to generate a kth discriminative crime element representationWherein k=1, 2,3,4, < >>Based on the action sequence->Context sentence representation and sentence-level context vector +>The invention calculates the attention weight vector of a sentence levelNamely:
wherein the method comprises the steps ofIs a trainable weight matrix of sentence-level attention mechanisms. Note that if sequence a k All agent actions in (1) are 0, then set +.>Then, the expression method of the kth crime element is calculated as:
wherein the method comprises the steps ofn f Refers to the number of sentences in the fact description. In order to obtain the discrimination result of the kth crime element, a multi-label classifier is used for calculating:
wherein the method comprises the steps ofAnd->Is a parameter of a discriminant classifier for the kth crime element.
4. Multitasking decision predictor
In order to accurately predict the decision result, the invention connects the representations of four crime elements as the representation of the fact description f, i.eBased on v f Generating a specific fact representation for the ith legal decision prediction subtask>Wherein->
Wherein the method comprises the steps ofAnd->Is a learnable parameter for the i-th legal decision prediction subtask. Sigma (·) represents a nonlinear activation function (in the present invention the activation function is relu). In order to obtain the prediction result of legal decision, the invention adopts a linear classifier:
wherein the method comprises the steps ofAnd->Is a learnable parameter for the i-th legal decision prediction subtask.
Experimental results on a real-world data set show that the method can effectively distinguish confusable crime facts and legal clauses, and improves accuracy of legal judgment prediction tasks.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
While the foregoing description of the embodiments of the present invention has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the invention, but rather, it is intended to cover all modifications or variations within the scope of the invention as defined by the claims of the present invention.
Claims (7)
1. A legal judgment prediction method based on crime elements is characterized in that: the method comprises the following steps:
carrying out fact description coding by using the layering Bi-LSTM, and generating sentence representation of the context; the specific process of carrying out fact description coding by using the layering Bi-LSTM comprises the following steps: using four hierarchical Bi-LSTM, each hierarchical Bi-LSTM having two levels of Bi-LSTM, word-level Bi-LSTM outputting a contextual word representation, computing a basic sentence representation by an attention mechanism, and using sentence-level Bi-LSTM outputting a contextual sentence representation;
extracting elements based on reinforcement learning, and selecting sentences containing crime elements; the specific process for extracting the elements based on reinforcement learning comprises the following steps: four agents are adopted to respectively select sentences containing criminals of different types of criminals including criminals, criminal targets, criminal intentions and criminal behaviors, and each agent adopts a random strategy to sample actions in each state;
generating a corresponding discriminating criminal element representation by fusing the contextual representations of the selected sentences; the specific process of the contextual representation of the selected sentence by fusion includes: fusing the context representations of sentences containing the crimes to generate a kth distinguished crimes representation, calculating attention weight vectors for one sentence level based on the action sequences, the context sentence representations and the sentence-level context vectors, and calculating the representation of the crimes based on the attention weight vectors;
and splicing criminal element representations for distinguishing criminals, criminal targets, criminal intentions and criminal behaviors, and performing multitask judgment and prediction on the formed spliced vectors to obtain a prediction judgment result.
2. The criminal element-based legal decision prediction method as described in claim 1, wherein: each fact description has a plurality of sentences, each sentence comprising a plurality of words, for each input sentence in the fact description, outputting a sequence of contextual word representations using word-level Bi-LSTM, calculating a word-level attention vector based on the sequence of contextual word representations and the word-level context vector, obtaining a basic representation of the sentence, calculating a basic sentence representation sequence based on the word-level attention mechanism, and using the sentence-level Bi-LSTM to calculate the sequence of contextual sentence representations.
3. The criminal element-based legal decision prediction method as described in claim 1, wherein: the method comprises the steps of measuring the extraction correctness of criminals by utilizing different delay rewards, providing a first delay reward for a kth agent by judging the difference between the kth criminals extracted by the agents and the real kth criminals, providing a second delay reward for all agents by calculating a prediction error of relative real labeling legal terms, and calculating the final reward of each agent by utilizing the sum of the two delay rewards.
4. The criminal element-based legal decision prediction method as described in claim 1, wherein: the specific process of carrying out multitask judgment and prediction on the formed spliced vector comprises the following steps: the representations of the four crime elements are linked together as representations of the fact descriptions, and a fact representation is generated for the respective legal decision prediction subtasks based on the representations of the fact descriptions.
5. A legal judgment prediction system based on crime elements is characterized in that: comprising the following steps:
a fact description encoder configured to encode the fact description using the hierarchical Bi-LSTM, generating a sentence representation of the context; the specific process of carrying out fact description coding by using the layering Bi-LSTM comprises the following steps: using four hierarchical Bi-LSTM, each hierarchical Bi-LSTM having two levels of Bi-LSTM, word-level Bi-LSTM outputting a contextual word representation, computing a basic sentence representation by an attention mechanism, and using sentence-level Bi-LSTM outputting a contextual sentence representation;
an element extractor configured to extract elements by reinforcement learning, and to pick out sentences containing crime elements; the specific process for extracting the elements based on reinforcement learning comprises the following steps: four agents are adopted to respectively select sentences containing criminals of different types of criminals including criminals, criminal targets, criminal intentions and criminal behaviors, and each agent adopts a random strategy to sample actions in each state;
a crime element discriminator configured to generate a corresponding discriminated crime element representation by fusing the contextual representations of the selected sentences; the specific process of the contextual representation of the selected sentence by fusion includes: fusing the context representations of sentences containing the crimes to generate a kth distinguished crimes representation, calculating attention weight vectors for one sentence level based on the action sequences, the context sentence representations and the sentence-level context vectors, and calculating the representation of the crimes based on the attention weight vectors;
the multi-task judgment predictor is configured to splice criminal element representations for distinguishing criminals, criminal targets, criminal intentions and criminal behaviors, and to perform multi-task judgment prediction on formed spliced vectors to obtain a prediction judgment result.
6. A computer-readable storage medium, characterized by: in which instructions are stored which are adapted to be loaded by a processor of a terminal device and to carry out the steps of a crime element based legal decision prediction method as claimed in any of claims 1-4.
7. A terminal device, characterized by: comprising a processor and a computer-readable storage medium, the processor configured to implement instructions; a computer readable storage medium for storing a plurality of instructions adapted to be loaded by a processor and to perform the steps of a crime element-based legal decision prediction method as recited in any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110632603.3A CN113487453B (en) | 2021-06-07 | 2021-06-07 | Legal judgment prediction method and system based on crime elements |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110632603.3A CN113487453B (en) | 2021-06-07 | 2021-06-07 | Legal judgment prediction method and system based on crime elements |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113487453A CN113487453A (en) | 2021-10-08 |
| CN113487453B true CN113487453B (en) | 2023-12-05 |
Family
ID=77934374
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110632603.3A Active CN113487453B (en) | 2021-06-07 | 2021-06-07 | Legal judgment prediction method and system based on crime elements |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113487453B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113505937B (en) * | 2021-07-26 | 2022-03-29 | 江西理工大学 | Multi-view encoder-based legal decision prediction system and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458373A (en) * | 2019-08-22 | 2019-11-15 | 电子科技大学成都学院 | A kind of method of crime prediction and system of the fusion of knowledge based map |
| CN110610005A (en) * | 2019-09-16 | 2019-12-24 | 哈尔滨工业大学 | Stealing crime auxiliary criminal investigation method based on deep learning |
| CN112116172A (en) * | 2020-09-30 | 2020-12-22 | 四川大学 | Criminal phase prediction method based on probability map model |
| CN112287689A (en) * | 2020-10-27 | 2021-01-29 | 山东省计算中心（国家超级计算济南中心） | Judicial second-examination case situation auxiliary analysis method and system |
| CN112329891A (en) * | 2020-11-27 | 2021-02-05 | 浙江大学 | Auxiliary case decision method, device and medium combining bidirectional attention and case decision logic |
| CN112508269A (en) * | 2020-12-03 | 2021-03-16 | 中国科学技术大学 | Legal decision prediction method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019023500A1 (en) * | 2017-07-26 | 2019-01-31 | The Trustees Of Dartmouth College | Computer-implemented perceptual apparatus |
-
2021
- 2021-06-07 CN CN202110632603.3A patent/CN113487453B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458373A (en) * | 2019-08-22 | 2019-11-15 | 电子科技大学成都学院 | A kind of method of crime prediction and system of the fusion of knowledge based map |
| CN110610005A (en) * | 2019-09-16 | 2019-12-24 | 哈尔滨工业大学 | Stealing crime auxiliary criminal investigation method based on deep learning |
| CN112116172A (en) * | 2020-09-30 | 2020-12-22 | 四川大学 | Criminal phase prediction method based on probability map model |
| CN112287689A (en) * | 2020-10-27 | 2021-01-29 | 山东省计算中心（国家超级计算济南中心） | Judicial second-examination case situation auxiliary analysis method and system |
| CN112329891A (en) * | 2020-11-27 | 2021-02-05 | 浙江大学 | Auxiliary case decision method, device and medium combining bidirectional attention and case decision logic |
| CN112508269A (en) * | 2020-12-03 | 2021-03-16 | 中国科学技术大学 | Legal decision prediction method and system |
Non-Patent Citations (1)
| Title |
|---|
| BLHNN: A Novel Charge Prediction Model Based on Bi-Attention LSTM-CNN Hybrid Neural Network;Jia Guo等;《IEEE》;第246-252页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113487453A (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112015859B (en) | Knowledge hierarchy extraction method and device for text, computer equipment and readable medium | |
| CN112732871B (en) | Multi-label classification method for acquiring client intention labels through robot induction | |
| CN107103363B (en) | A kind of construction method of the software fault expert system based on LDA | |
| CN113326380B (en) | Equipment measurement data processing method, system and terminal based on deep neural network | |
| CN113254610A (en) | Multi-round conversation generation method for patent consultation | |
| CN113052149A (en) | Video abstract generation method and device, computer equipment and medium | |
| CN113435998B (en) | Loan overdue prediction method and device, electronic equipment and storage medium | |
| Liu et al. | Saliency as evidence: Event detection with trigger saliency attribution | |
| CN111985207B (en) | Access control policy acquisition method and device and electronic equipment | |
| AU2023200391A1 (en) | Systems and methods for determining structured proceeding outcomes | |
| CN113487453B (en) | Legal judgment prediction method and system based on crime elements | |
| CN116150367A (en) | Emotion analysis method and system based on aspects | |
| CN113836934A (en) | Text classification method and system based on label information enhancement | |
| CN117151222B (en) | Domain knowledge guided emergency case entity attribute and relation extraction method thereof, electronic equipment and storage medium | |
| CN111259673B (en) | Legal decision prediction method and system based on feedback sequence multitask learning | |
| CN117648950A (en) | Training method and device for neural network model, electronic equipment and storage medium | |
| CN115357712A (en) | Aspect level emotion analysis method and device, electronic equipment and storage medium | |
| He et al. | Learning label-relational output structure for adaptive sequence labeling | |
| CN113505937A (en) | Multi-view encoder-based legal decision prediction system and method | |
| CN111949791B (en) | Text classification method, device and equipment | |
| CN116610783B (en) | Service optimization method based on artificial intelligent decision and digital online page system | |
| CN116227598B (en) | Event prediction method, device and medium based on dual-stage attention mechanism | |
| US11928466B2 (en) | Distributed representations of computing processes and events | |
| CN114169310B (en) | Reading and understanding answer generation method and device | |
| CN113342974B (en) | Method, device and equipment for identifying overlapping relationship of network security entities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |