CN109919368A - A kind of law article recommendation forecasting system and method based on associated diagram - Google Patents
A kind of law article recommendation forecasting system and method based on associated diagram Download PDFInfo
- Publication number
- CN109919368A CN109919368A CN201910142623.5A CN201910142623A CN109919368A CN 109919368 A CN109919368 A CN 109919368A CN 201910142623 A CN201910142623 A CN 201910142623A CN 109919368 A CN109919368 A CN 109919368A
- Authority
- CN
- China
- Prior art keywords
- law article
- law
- classifier
- node
- associated diagram
- 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.)
- Granted
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application provides a kind of law article recommender system and method based on associated diagram, it include: to obtain merit description, it is related to each law article and the acquisition module of adduction relationship therebetween of law, utilize the associated diagram constructing module of adduction relationship construction associated diagram between law article, construct learning model, extract the extraction of semantics module for obtaining that the main semantic information of text is described comprising merit, each of corresponding associated diagram joint structure classifier prediction node correspondent method strip label is the classifier constructing module of genuine possibility and the next step flow direction of the node, utilize the training module of this network model being made of classifier of the optimization algorithm training of deep learning.This system is applied to law article and recommends task, can carry out multi-tag prediction using adduction relationship between law article, promote the accuracy rate of law article recommended work.
Description
Technical field
The invention belongs to Text Classification field, in particular to a kind of law article based on associated diagram recommend forecasting system and
Method.
Background technique
In recent years, with the propulsion of " wisdom law court ", the informatization of people's court is constantly improve, and case trial is more next
More come into the open, transparence.Ending in December, 2018, Chinese judgement document's net has included more than 59,000,000 judgement documents, wherein
Present case information abundant and trial situation.Also contain huge value among the judicial data of these magnanimity.
During case trial, judge generally requires to consult the trial situation of relevant legal articles and previous similar case,
And case is analyzed in conjunction with these contents, this part work has tended to take up the plenty of time of case trial.Although borrowing
Help existing information-based means, retrieval work can be completed by computer, but more cumbersome browsed, analyzed work still
It can only be by manually completing.
With the development of machine learning techniques, completing these work using previous judicial data becomes possibility.It utilizes
Deep neural network, computer analyze it, Xiang Faguan intelligent recommendation case correlation method it will be appreciated that case content
Item greatly reduces the workload of judge, promotes the efficiency of case trial.
Law article recommended work is substantially applied to the text classification problem of judicial domain, often uses convolutional Neural
The modes such as network, Recognition with Recurrent Neural Network first extract the feature implied in text, then are classified with classifier to these features.But
Existing classification method is most of only to support single labeling, i.e. a text to be only capable of corresponding on a classification or label, and
Existing some multi-tag classification methods tend not to although multiple labels can be stamped for a text in view of this
Relevance between a little labels.It there is inclusion relation for example, label as " cuisines " and " Sichuan cuisine ", between them.Obviously, this
The value that the incidence relation of sample will cause a label influences the value of other labels, therefore, is closed using the association between label
System can assist the prediction of other label values using fixed label value, effectively promote the effect of multi-tag prediction.
Specific to the problem of in law article recommendation, this task itself is a multi-tag classification, a merit is possible to
It is related to multiple related law articles, and can also has association between law article.For example, often have among criminal law " ..., according to this law
The regulation of ×× item is punished ".Therefore, present applicant proposes a kind of multi-tags based on associated diagram for recommending task applied to law article
Classification method and device are solved previous methods and device cannot be fine to be recommended using the incidence relation auxiliary law article between law article
Ground carries out law article recommendation to the merit for being involved in the problems, such as a plurality of law.
Summary of the invention
In order to overcome the disadvantages of the above prior art, the purpose of the present invention is to provide a kind of law articles based on associated diagram to push away
Forecasting system and method are recommended, it is mutually indepedent when solving traditional each law article of law article recommended method prediction, it cannot consider to close between law article
The problem of connection property.
To achieve the goals above, the technical solution adopted by the present invention is that:
A kind of law article recommendation forecasting system based on associated diagram, comprising:
Module is obtained, the adduction relationship between all law articles and law article for obtaining merit description, relevant law；
Associated diagram constructing module, for constructing a directed acyclic graph as association using the adduction relationship between the law article
Scheme, each node indicates that a law article, each edge indicate the adduction relationship between law article in figure, and the direction on side is the side of adduction relationship
To；
Extraction of semantics module is obtained for describing the extraction that building learning model carries out feature to merit comprising main language
The feature vector of adopted information；
Classifier constructing module, including label prediction unit and flow to predicting unit, it is every in corresponding constructed associated diagram
One node, constructs a classifier, and label prediction unit carries out two classification to the corresponding law article of the node, predict the law article with
The relevance scores of merit flow to predicting unit and predict the flow direction of the node next step if the node has downstream node,
According to prediction result using law article and the relevance scores of merit as the input of the classifier of respective downstream nodes in associated diagram,
That is, the input of each classifier includes at least the extracted feature vector of extraction of semantics module, and in associated diagram middle and upper reaches node
When corresponding classifier has incoming, while the input being passed to including the corresponding classifier of upstream node；
Training module obtains the network model being made of classifier, netted mould after constructing to all classifiers
Each classifier of type will correspond to law article to it and predict, it is determined whether recommend the law article, training module is netted for calculating
The loss function of model entirety, and using this model of gradient back-propagation method even depth study optimization algorithm training, make model
According to training data adjusting parameter, model can be used to law article recommended work after training.
The module that obtains collects the merit data for being used for model training first, in the judgement document including each case
Fact description and its relevant law being related to, and determine the set of the corresponding all law articles of relevant law of merit, then determine method
Adduction relationship in item set between each law article, i.e., influence of the value of a certain method strip label to other law article label values.
The extraction of semantics module describes text to merit, and building deep learning model carries out feature extraction, included
The feature vector of the main semantic information of text, feature vector can indicate the information that text includes in different aspect.
The label prediction unit carries out two classification to the corresponding law article of node where classifier, uses machine learning method
Predict a possibility that law article is related to merit；It is described to flow to predicting unit using machine learning method to the node next step
Flow direction is predicted, and the law article prediction result that the label prediction unit obtains is sent to the next step stream that this unit is predicted
To input as the classifier of respective downstream nodes in associated diagram.
The present invention also provides a kind of, and the law article based on associated diagram recommends prediction technique, comprising the following steps:
Step 1: obtain merit description, relevant law all law articles and law article between adduction relationship；
Step 2: constructing a directed acyclic graph as associated diagram, in figure using the adduction relationship between law article described in step 1
Each node indicates that a law article, each edge indicate the adduction relationship between law article, and the direction on side is the direction of adduction relationship；
Step 3: merit being described, building deep learning model carries out feature extraction, obtains comprising main semantic information
Feature vector；
Step 4: each of corresponding associated diagram node constructs a classifier, carries out to the corresponding law article of the node
Two classification, predict the relevance scores of the law article and merit, if the node has downstream node, to the flow direction of the node next step into
Row prediction, according to flowing to prediction result using obtained law article and the relevance scores of merit as respective downstream section in associated diagram
Point classifier input, that is, the input of each classifier include at least the extracted feature vector of extraction of semantics module, and
When the corresponding classifier of associated diagram middle and upper reaches node has incoming, while the input being passed to including the corresponding classifier of upstream node；
Step 5: after to all classifiers construction, obtaining the network model being made of classifier, network model
Each classifier will correspond to law article to it and predict, it is determined whether recommend the law article, the loss function of computation model entirety,
And using this network model of the optimization algorithm of deep learning training, model can be used to law article recommended work after training.
The classifier is neural network or conventional machines learning model.
The input of the classifier includes: to extract the classifier biography of obtained merit Text eigenvector and upstream node
The input entered, output include: the input of the prediction result of the corresponding method strip label of node and the classifier of downstream node.
Compared with prior art, the present invention can be first laggard to each law article and the correlation of merit according to law article adduction relationship
Row prediction promotes law article and recommends accuracy rate to overcome the shortcomings of that existing law article recommended method cannot consider law article adduction relationship.
Detailed description of the invention
Fig. 1 is this system structural schematic diagram.
Fig. 2 is the associated diagram pattern example that the associated diagram constructing module of this system constructs.
Fig. 3 is the associated diagram part constructed in embodiment.
Fig. 4 is the structural schematic diagram of the classifier constructing module of this system.
Fig. 5 is the classifier that constructs and its to output and input schematic diagram in the classifier constructing module of this system.
Fig. 6 is the classifier network model part constructed in embodiment.
Fig. 7 is law article recommended method flow chart provided by the present application.
Specific embodiment
The embodiment that the present invention will be described in detail with reference to the accompanying drawings and examples.
Referring to Fig. 1, this application provides a kind of, and the law article based on associated diagram recommends forecasting system, comprising:
Module 101 is obtained, for obtaining merit description, all law articles of relevant law and adduction relationship therebetween.First
The merit data for being used for model training are collected, true description and its correlation method being related in the judgement document including each case
Rule, and determine the set of the corresponding all law articles of relevant law of merit.By taking criminal case as an example, for the merit by criminal case
It is related to the task of law article for description prediction, and all law articles in criminal law are law article set, each merit is described as one
Merit data correspond to several law articles in criminal law.On the other hand, it is also necessary to determine that the reference in law article set between each law article is closed
System, i.e., influence of the value of a certain method strip label to other law article label values.Still by taking criminal case as an example, criminal law the 260th
Five are related to robber and connect Telecom Facilities, and handled according to the 264th article (general larceny), therefore, if merit relates to
And the 265th article, then can also be related to the 264th article, i.e., the 264th article of the certain journey of label value
The 265th article of label value is depended on degree.
Associated diagram constructing module 102.A directed acyclic graph is constructed using adduction relationship between obtaining 101 gained law article of module
As associated diagram, associated diagram pattern is referring to fig. 2.Each node indicates that a law article, each edge indicate the reference between law article in figure
Relationship, the direction of the direction on side with adduction relationship.Each node can draw side and be connected to multiple nodes (claiming downstream node),
Side connection can be drawn by multiple nodes (claiming upstream node).In upper an example, " the 264th article of label value is to a certain degree
The upper label value dependent on the 265th article " can construct structure as shown in Figure 3, wherein dotted line be it is that may be present with
The association of other nodes.
Extraction of semantics module 103.Text is described for merit, constructs Recognition with Recurrent Neural Network, convolutional neural networks even depth
Learning model carries out the extraction of feature, obtains the feature vector comprising the main semantic information of text.Feature vector can indicate text
This information for including in different aspect.A variety of model realizations can be used in this module, therefore not to the study mould for obtaining feature vector
Type defines.
Classifier constructing module 104, each of corresponding constructed associated diagram node, constructs a classifier, each
The input of classifier is the classification corresponding with its upstream node in associated diagram of the extracted feature vector of extraction of semantics module
The incoming input of device (may include the incoming input of multiple upstream nodes, it is also possible to for sky), classifier is to the corresponding method of the node
Item carries out two classification, predicts the relevance scores of the law article and merit, will also be right if node where classifier has downstream node
The flow direction of the node next step is predicted, according to prediction result using obtained law article and the relevance scores of merit as pass
Join the input of the classifier of respective downstream nodes in figure；
After constructing to all classifiers, a network model being made of classifier is can be obtained in training module 105,
Each classifier of network model will correspond to law article to it and predict, it is determined whether recommend the law article, training module is based on
The loss function of model entirety is calculated, and using this model of gradient back-propagation method even depth study optimization algorithm training, makes mould
Type is according to training data adjusting parameter, and model can be used to law article recommended work after training.
Further, referring to fig. 4, the classifier constructing module 104 includes:
Label prediction unit 201 carries out two classification for the law article to classifier corresponding node, predicts the law article and merit
A possibility that related, the relevance scores as law article and merit.As shown in figure 5, there are two inputs for classifier, it is respectively semantic
The incoming input of the feature vector classifier corresponding with its upstream node in associated diagram that extraction module 103 extracts (can wrap
The input being passed to containing multiple upstream nodes, it is also possible to for sky), label prediction unit carries out two to the corresponding law article label of the node
Classification, predicting that the node corresponds to law article is genuine possibility, as output.
Flow to predicting unit 202, predicted for the flow direction to the node next step, and to predict flow direction carry out it is defeated
Out.Flow direction prediction, which refers to, judges all downstream nodes of the node, predicts whether to export node, all predictions
Node for "Yes" is to flow to prediction result.When output, law article and the correlation of merit that label prediction unit 201 is predicted
Score is as the input for flowing to the corresponding classifier of all nodes in prediction result.Still with the introduction of associated diagram constructing module 102
In citing be illustrated, referring to Fig. 6.264th article and the 265th article is respectively a section in associated diagram
Point, respectively they construct a classifiers, two classifiers be respectively used to prediction merit whether with the 264th article and
265th article of correlation, while the 265th article of corresponding classifier can predict its next step flow direction, i.e., whether
Result of merit whether related to the 265th article is input to the 264th article of corresponding classifier.It is predicting
When, it is flowed to first by the 264th article of corresponding classifier prediction law article and true relevance scores and in next step, if stream
The 260th can be used as comprising the relevance scores of the 265th article, the 264th article and merit to prediction result
The input of five corresponding classifiers, it is on the contrary then will not.
From the above technical scheme, the law article based on associated diagram that this application provides a kind of recommends forecasting system, first really
Adduction relationship between fixed each law article, then successively predicts each law article label value using it, realizes that more law articles are predicted jointly,
Overcome the shortcomings of that existing law article recommended method and device cannot consider law article adduction relationship, promotes the accuracy rate that law article is recommended.
As shown in fig. 7, this application provides a kind of, the law article based on associated diagram recommends prediction technique, comprising the following steps:
Step 301: collecting the merit data for being used for model training first, the fact in the judgement document including each case
Description and its relevant law being related to, and determine the set of the corresponding all law articles of relevant law of merit.On the other hand, it is also necessary to
Determine the adduction relationship in law article set between each law article, i.e., influence of the value of a certain method strip label to other law article label values.
Step 302: constructing a directed acyclic graph as associated diagram using the adduction relationship between law article described in step 301.
Each node indicates a law article in figure, and each edge indicates the adduction relationship between law article, the direction of the direction on side with adduction relationship.
Each node can draw side and be connected to multiple nodes, and side connection can also be drawn by multiple nodes.
Step 303: text being described for merit, constructs Recognition with Recurrent Neural Network, convolutional neural networks even depth learning model
The extraction for carrying out feature, obtains the feature vector comprising the main semantic information of text.Feature vector can indicate text in difference
The information that aspect includes.This method does not define the learning model for obtaining feature vector.
Step 304: each of corresponding associated diagram node uses neural network or conventional machines learning method construction one
A classifier, the law article label for indicating the node are predicted.It is not construed as limiting in specific implementation this method of classifier.
There are two inputs, the respectively feature vector of step 303 extraction and all nodes by being relied in associated diagram for each classifier
The incoming input of corresponding classifier (input that can be passed to comprising multiple upstream nodes, it is also possible to for sky), classifier pair
The corresponding law article label of the node carries out two classification, and predicting that the node corresponds to law article is genuine possibility, as output.
Step 305: another output of classifier is the flow direction of the node next step, if node where classifier has downstream
Node, the obtained Tag Estimation result of step 304 are input to respective downstream section in associated diagram according to the flow direction that classifier is predicted
The corresponding classifier of point.The flow direction of prediction may include several nodes, these nodes will all obtain the method for present node prediction
Strip label value.If prediction flow direction is sky, the input that thus node obtains of the corresponding classifier of all nodes in downstream is sky.
Step 306: after to all classifiers construction, a network model being made of classifier can be obtained, it is netted
Each classifier of model will correspond to law article to it and predict, it is determined whether recommend the law article, i.e. whether merit is related to the method
Item.Using this network model of gradient back-propagation method even depth study optimization algorithm training, make model according to training data tune
Whole inner parameter, model can be used to law article recommended work after training.
From the above technical scheme, the law article based on associated diagram that this application provides a kind of recommends prediction technique, first really
Adduction relationship between fixed each law article, then successively predicts each law article label value using it, realizes that more law articles are predicted jointly,
Overcome the shortcomings of that existing law article recommended method and device cannot consider law article adduction relationship, promotes the accuracy rate that law article is recommended.
The embodiment of System and method for is substantially similar in this specification, and related place can be cross-referenced.
The above, the only specific embodiment in the present invention, but scope of protection of the present invention is not limited thereto, appoints
What is familiar with the people of the technology within the technical scope disclosed by the invention, it will be appreciated that expects transforms or replaces, and should all cover
Within scope of the invention, therefore, the scope of protection of the invention shall be subject to the scope of protection specified in the patent claim.
Claims (7)
1. a kind of law article based on associated diagram recommends forecasting system characterized by comprising
Module is obtained, the adduction relationship between all law articles and law article for obtaining merit description, relevant law；
Associated diagram constructing module, for constructing a directed acyclic graph as associated diagram using the adduction relationship between the law article,
Each node indicates that a law article, each edge indicate the adduction relationship between law article in figure, and the direction on side is the direction of adduction relationship；
Extraction of semantics module is obtained for describing the extraction that building learning model carries out feature to merit comprising main semantic letter
The feature vector of breath；
Classifier constructing module, including label prediction unit and flow to predicting unit, each of corresponding constructed associated diagram
Node, constructs a classifier, and label prediction unit carries out two classification to the corresponding law article of the node, predicts the law article and merit
Relevance scores flow to predicting unit if the node has downstream node and the flow direction of the node next step predicted, according to
Prediction result is using law article and the relevance scores of merit as the input of the classifier of respective downstream nodes in associated diagram, that is, every
The input of a classifier includes at least the extracted feature vector of extraction of semantics module, and corresponding in associated diagram middle and upper reaches node
When classifier has incoming, while the input being passed to including the corresponding classifier of upstream node；
Training module obtains the network model being made of classifier after constructing to all classifiers, network model
Each classifier will correspond to law article to it and predict, it is determined whether recommend the law article, training module is for calculating network model
Whole loss function, and using this model of gradient back-propagation method even depth study optimization algorithm training, make model according to
Training data adjusting parameter, model can be used to law article recommended work after training.
2. the law article based on associated diagram recommends forecasting system according to claim 1, which is characterized in that the acquisition module is first
The merit data for being used for model training are first collected, the true description in the judgement document including each case is related to related to it
Law, and determine the set of the corresponding all law articles of relevant law of merit, then determine the reference in law article set between each law article
Relationship, i.e., influence of the value of a certain method strip label to other law article label values.
3. the law article based on associated diagram recommends forecasting system according to claim 1, which is characterized in that the extraction of semantics mould
Block describes text to merit, and building deep learning model carries out feature extraction, obtains the feature comprising the main semantic information of text
Vector, feature vector can indicate the information that text includes in different aspect.
4. the law article based on associated diagram recommends forecasting system according to claim 1, which is characterized in that the Tag Estimation list
Member carries out two classification to the corresponding law article of node where classifier, predicts that the law article is relevant to merit using machine learning method
Possibility；The predicting unit that flows to predicted using flow direction of the machine learning method to the node next step, and will be described
The law article prediction result that label prediction unit obtains is sent to the next step flow direction that this unit is predicted, under corresponding in associated diagram
Swim the input of the classifier of node.
5. a kind of law article based on associated diagram recommends prediction technique, which comprises the following steps:
Step 1: obtain merit description, relevant law all law articles and law article between adduction relationship；
Step 2: constructing a directed acyclic graph using the adduction relationship between law article described in step 1 and be used as associated diagram, in figure each
Node indicates that a law article, each edge indicate the adduction relationship between law article, and the direction on side is the direction of adduction relationship；
Step 3: merit being described, building deep learning model carries out feature extraction, obtains the spy comprising main semantic information
Levy vector；
Step 4: each of corresponding associated diagram node constructs a classifier, carries out two points to the corresponding law article of the node
Class predicts the relevance scores of the law article and merit, if the node has downstream node, carries out to the flow direction of the node next step pre-
It surveys, according to flowing to prediction result using obtained law article and the relevance scores of merit as respective downstream nodes in associated diagram
The input of classifier, that is, the input of each classifier includes at least the extracted feature vector of extraction of semantics module, and is being associated with
When the corresponding classifier of figure middle and upper reaches node has incoming, while the input being passed to including the corresponding classifier of upstream node；
Step 5: after to all classifiers construction, obtain the network model being made of classifier, network model it is every
A classifier will correspond to law article to it and predict, it is determined whether recommend the law article, the loss function of computation model entirety, and benefit
With this network model of the optimization algorithm training of deep learning, model can be used to law article recommended work after training.
6. the law article based on associated diagram recommends prediction technique according to claim 1, which is characterized in that the classifier is mind
Through network or conventional machines learning model.
7. law article based on associated diagram recommends prediction technique according to claim 1, which is characterized in that the classifier it is defeated
Entering includes: the input extracting the classifier of obtained merit Text eigenvector and upstream node and being passed to, and output includes: node
The input of the classifier of the prediction result and downstream node of corresponding method strip label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910142623.5A CN109919368B (en) | 2019-02-26 | 2019-02-26 | Law recommendation prediction system and method based on association graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910142623.5A CN109919368B (en) | 2019-02-26 | 2019-02-26 | Law recommendation prediction system and method based on association graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109919368A true CN109919368A (en) | 2019-06-21 |
| CN109919368B CN109919368B (en) | 2020-11-17 |
Family
ID=66962370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910142623.5A Active CN109919368B (en) | 2019-02-26 | 2019-02-26 | Law recommendation prediction system and method based on association graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109919368B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110737781A (en) * | 2019-10-21 | 2020-01-31 | 南京大学 | law and fact relation calculation method based on multi-layer knowledge |
| CN110750635A (en) * | 2019-10-21 | 2020-02-04 | 南京大学 | Joint deep learning model-based law enforcement recommendation method |
| CN110851584A (en) * | 2019-11-13 | 2020-02-28 | 成都华律网络服务有限公司 | Accurate recommendation system and method for legal provision |
| CN110941702A (en) * | 2019-11-26 | 2020-03-31 | 北京明略软件系统有限公司 | Retrieval method and device for laws and regulations and laws and readable storage medium |
| CN111583072A (en) * | 2020-04-27 | 2020-08-25 | 北京北大软件工程股份有限公司 | Judgment method for upper and lower relation of law bar and processing terminal |
| CN112163152A (en) * | 2020-09-29 | 2021-01-01 | 信阳农林学院 | Accurate recommendation system for legal provision |
| CN113377944A (en) * | 2020-12-02 | 2021-09-10 | 中国司法大数据研究院有限公司 | Case feature extraction and law enforcement recommendation method and device based on multiple tasks |
| CN113792127A (en) * | 2021-09-15 | 2021-12-14 | 平安国际智慧城市科技股份有限公司 | Big data-based law identification method and device, electronic equipment and medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204366A (en) * | 2016-07-22 | 2016-12-07 | 三峡大学 | A kind of big data management system of law based on fuzzy reasoning |
| CN106952193A (en) * | 2017-03-23 | 2017-07-14 | 北京华宇信息技术有限公司 | A kind of criminal case aid decision-making method based on fuzzy depth belief network |
| CN107577785A (en) * | 2017-09-15 | 2018-01-12 | 南京大学 | A kind of level multi-tag sorting technique suitable for law identification |
| CN108073673A (en) * | 2017-05-15 | 2018-05-25 | 北京华宇元典信息服务有限公司 | A kind of legal knowledge map construction method, apparatus, system and medium based on machine learning |
| CN108304386A (en) * | 2018-03-05 | 2018-07-20 | 上海思贤信息技术股份有限公司 | A kind of logic-based rule infers the method and device of legal documents court verdict |
| CN109241285A (en) * | 2018-08-29 | 2019-01-18 | 东南大学 | A kind of device of the judicial decision in a case of auxiliary based on machine learning |
| CN109299753A (en) * | 2018-12-11 | 2019-02-01 | 济南浪潮高新科技投资发展有限公司 | A kind of integrated learning approach and system for Law Text information excavating |
-
2019
- 2019-02-26 CN CN201910142623.5A patent/CN109919368B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106204366A (en) * | 2016-07-22 | 2016-12-07 | 三峡大学 | A kind of big data management system of law based on fuzzy reasoning |
| CN106952193A (en) * | 2017-03-23 | 2017-07-14 | 北京华宇信息技术有限公司 | A kind of criminal case aid decision-making method based on fuzzy depth belief network |
| CN108073673A (en) * | 2017-05-15 | 2018-05-25 | 北京华宇元典信息服务有限公司 | A kind of legal knowledge map construction method, apparatus, system and medium based on machine learning |
| CN107577785A (en) * | 2017-09-15 | 2018-01-12 | 南京大学 | A kind of level multi-tag sorting technique suitable for law identification |
| CN108304386A (en) * | 2018-03-05 | 2018-07-20 | 上海思贤信息技术股份有限公司 | A kind of logic-based rule infers the method and device of legal documents court verdict |
| CN109241285A (en) * | 2018-08-29 | 2019-01-18 | 东南大学 | A kind of device of the judicial decision in a case of auxiliary based on machine learning |
| CN109299753A (en) * | 2018-12-11 | 2019-02-01 | 济南浪潮高新科技投资发展有限公司 | A kind of integrated learning approach and system for Law Text information excavating |
Non-Patent Citations (3)
| Title |
|---|
| CHAOJUN XIAO ET AL: ""CAIL2018: A Large-Scale Legal Dataset for Judgment Prediction"", 《ARXIV》 * |
| HAOXI ZHONG ET AL: ""Legal Judgment Prediction via Topological Learning"", 《EMNLP2018》 * |
| 邓文超: ""基于深度学习的司法智能研究"", 《中国优秀硕士学位论文全文数据库 社会科学Ⅰ辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110737781A (en) * | 2019-10-21 | 2020-01-31 | 南京大学 | law and fact relation calculation method based on multi-layer knowledge |
| CN110750635A (en) * | 2019-10-21 | 2020-02-04 | 南京大学 | Joint deep learning model-based law enforcement recommendation method |
| CN110750635B (en) * | 2019-10-21 | 2023-12-15 | 南京大学 | French recommendation method based on joint deep learning model |
| CN110851584A (en) * | 2019-11-13 | 2020-02-28 | 成都华律网络服务有限公司 | Accurate recommendation system and method for legal provision |
| CN110851584B (en) * | 2019-11-13 | 2023-12-15 | 成都华律网络服务有限公司 | Legal provision accurate recommendation system and method |
| CN110941702A (en) * | 2019-11-26 | 2020-03-31 | 北京明略软件系统有限公司 | Retrieval method and device for laws and regulations and laws and readable storage medium |
| CN111583072A (en) * | 2020-04-27 | 2020-08-25 | 北京北大软件工程股份有限公司 | Judgment method for upper and lower relation of law bar and processing terminal |
| CN111583072B (en) * | 2020-04-27 | 2023-11-07 | 北京北大软件工程股份有限公司 | Legal upper and lower relationship judging method and processing terminal |
| CN112163152A (en) * | 2020-09-29 | 2021-01-01 | 信阳农林学院 | Accurate recommendation system for legal provision |
| CN113377944A (en) * | 2020-12-02 | 2021-09-10 | 中国司法大数据研究院有限公司 | Case feature extraction and law enforcement recommendation method and device based on multiple tasks |
| CN113792127A (en) * | 2021-09-15 | 2021-12-14 | 平安国际智慧城市科技股份有限公司 | Big data-based law identification method and device, electronic equipment and medium |
| CN113792127B (en) * | 2021-09-15 | 2023-12-26 | 平安国际智慧城市科技股份有限公司 | Rule recognition method and device based on big data, electronic equipment and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109919368B (en) | 2020-11-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109919368A (en) | A kind of law article recommendation forecasting system and method based on associated diagram | |
| CN108573411B (en) | Mixed recommendation method based on deep emotion analysis and multi-source recommendation view fusion of user comments | |
| CN109471938A (en) | A kind of file classification method and terminal | |
| CN107330396A (en) | A kind of pedestrian's recognition methods again based on many attributes and many strategy fusion study | |
| Mayatopani et al. | CLASSIFICATION OF VEHICLE TYPES USING BACKPROPAGATION NEURAL NETWORKS WITH METRIC AND ECCENTRICITY PARAMETERS | |
| CN109918560A (en) | A kind of answering method and device based on search engine | |
| CN109815952A (en) | Brand name recognition methods, computer installation and computer readable storage medium | |
| CN104199874A (en) | Webpage recommendation method based on user browsing behaviors | |
| CN109785629A (en) | A kind of short-term traffic flow forecast method | |
| CN109785064A (en) | A kind of mobile e-business recommended method and system based on Multi-source Information Fusion | |
| CN102165464A (en) | Method and system for automated annotation of persons in video content | |
| CN111079594B (en) | Video action classification and identification method based on double-flow cooperative network | |
| CN109214263A (en) | A kind of face identification method based on feature multiplexing | |
| Ahmadi et al. | Intelligent classification of web pages using contextual and visual features | |
| CN101470731A (en) | Personalized web page filtering method | |
| US20220292328A1 (en) | Convolutional artificial neural network based recognition system in which registration, search, and reproduction of image and video are divided between and performed by mobile device and server | |
| CN110188596A (en) | Monitor video pedestrian real-time detection, Attribute Recognition and tracking and system based on deep learning | |
| CN107169106A (en) | Video retrieval method, device, storage medium and processor | |
| CN106326451B (en) | A kind of webpage heat transfer agent block decision method of view-based access control model feature extraction | |
| CN113312474A (en) | Similar case intelligent retrieval system of legal documents based on deep learning | |
| CN105653563B (en) | The method and relevant apparatus of blacklist and white list are updated to control method, the dynamic of webpage capture | |
| CN110297986A (en) | A kind of Sentiment orientation analysis method of hot microblog topic | |
| CN102163285A (en) | Cross-domain video semantic concept detection method based on active learning | |
| CN107015965A (en) | A kind of Chinese text sentiment analysis device and method | |
| CN102063497B (en) | Open type knowledge sharing platform and entry processing method thereof |
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 |