CN106296495A - The Forecasting Methodology of a kind of lawsuit result and system - Google Patents
The Forecasting Methodology of a kind of lawsuit result and system Download PDFInfo
- Publication number
- CN106296495A CN106296495A CN201610647687.7A CN201610647687A CN106296495A CN 106296495 A CN106296495 A CN 106296495A CN 201610647687 A CN201610647687 A CN 201610647687A CN 106296495 A CN106296495 A CN 106296495A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- training
- document
- inference
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 131
- 238000013528 artificial neural network Methods 0.000 claims abstract description 103
- 238000012360 testing method Methods 0.000 claims abstract description 83
- 230000008569 process Effects 0.000 claims abstract description 32
- 239000000284 extract Substances 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims description 19
- 230000007935 neutral effect Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 238000012937 correction Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 6
- 239000000203 mixture Substances 0.000 claims description 5
- 238000013480 data collection Methods 0.000 claims description 3
- 238000004148 unit process Methods 0.000 claims description 3
- 241001269238 Data Species 0.000 claims description 2
- 230000008439 repair process Effects 0.000 claims description 2
- 238000013507 mapping Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 230000002123 temporal effect Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 3
- 210000005036 nerve Anatomy 0.000 description 3
- 230000008933 bodily movement Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000004218 nerve net Anatomy 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 230000008094 contradictory effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
Classifications
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/046—Forward inferencing; Production systems
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Technology Law (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides Forecasting Methodology and the system of a kind of lawsuit result, described method includes: set up the knowledge reasoning storehouse of predetermined theme；Knowledge reasoning storehouse is used for storing rule of inference, the multiple inference step of rule of inference；Knowledge reasoning storehouse based on predetermined theme, builds forecast model；Wherein, the corresponding inference step of the neural network block of each forecast model；The corresponding data type of inference data needed for each inference step, the input type of each data type correspondence neural network block；Process judgement document corresponding to predetermined theme and obtain training sample data set, train and revise forecast model；Obtain testing data, extract the data pair with attribute and property value based on predefined feature dictionary, form testing data characteristic information set；Based on testing data characteristic information set, it is judged that applicable theme the input prediction model of testing data are predicted.The present invention can instruct the result of people's predicted method rule case, practical.
Description
Technical field
The present invention relates to electric powder prediction based on neutral net, be specifically related to a kind of lawsuit result Forecasting Methodology and
System.
Background technology
Carry along with quickly the increasing of the new method of country, the fast propagation of legal information make the legal consciousness of people be commonly available
Height, legal requirements constantly expands also with the raising of quality of life.But, the general knowledge of the professional and user of law is shallow but
All the time annoying general public to the study of law and use, and asymmetric, the information source of legal industry information mixed and disorderly and
The feature of provision process legal issue also considerably increases general public and uses the difficulty of legal weapon right-safeguarding, example according to law
As, it's lawsuit effect duration the understanding of lawsuit effect duration causes past some legal requirements still look for requiring efforts owing to lacking
Lawyer seeks advice from；Owing to evidence dynamics is totally insufficient to support that demand is the most still spending time and efforts in prosecution etc..In view of this,
A kind of method and system that can instruct people that legal case is entered prediction becomes current urgent needs.
Summary of the invention
Embodiments provide Forecasting Methodology and the system of a kind of lawsuit result, on the one hand based on China's provision genealogy of law
Feature, from legal provision, the trial rule of technicalization case, builds Analysis of Knowledge Bases Reasoning storehouse, is used for instructing neutral net mould
Type prediction process；On the other hand, based on the big data of real lawsuit, analysis obtains at evidence similar with demand, and evidence proves
Basic case there is similarity, the case trial thinking feature in same region；Under Analysis of Knowledge Bases Reasoning instructs, structure
Building the legal case result Inference Forecast model of multiple neural network block, active client initial conditions is trained prediction may
The trial result obtained, makes up the weak tendency in domestic consumer's general knowledge of laws or analysis ability to a certain extent, strengthens from bodily movement of practising Wushu
The degree of understanding of rule demand.
According to an aspect of the present invention, the invention provides the Forecasting Methodology of a kind of lawsuit result, including:
Obtain testing data；
Based on predefined feature dictionary, extract number to be measured corresponding with described predefined Feature Words in described testing data
According to right, to form testing data characteristic information set；Preferably, described testing data characteristic information set at least include by case by
Three groups of testing datas pair that information, demand information and evidence set are constituted, often organize described testing data to a corresponding data class
Type；
Based on described testing data characteristic set, determine the applicable theme of described testing data；
Described testing data characteristic information set input is predicted with the described forecast model being suitable for theme corresponding,
To predicting the outcome of described testing data.
Preferably, in the step that described testing data characteristic information set input multiple neural network forecast model is predicted
Before Zhou, also include:
Knowledge reasoning storehouse based on the predetermined theme set up, builds described multiple neural network block forecast model.
Preferably, described knowledge reasoning storehouse based on the predetermined theme set up, build described multiple neural network block prediction mould
The step of type includes:
Set up the knowledge reasoning storehouse of predetermined theme；Wherein, described predetermined theme includes that case is by information and regional information；Described
Knowledge reasoning stock contains at least one rule of inference, and described rule of inference is by carrying out the provision corresponding with described theme
Extraction process and logical process obtain, and each described rule of inference at least includes multiple inference steps with logical relation；
Based on described knowledge reasoning storehouse, build described multiple neural network block forecast model；
Wherein it is preferred to, each neural network block of described forecast model is corresponding with inference step at least one described；Each
The corresponding data type of the inference data of described inference step, each described data type is the data of a neural network block
Input type；Wherein, the most each described inference step of described inference data makes inferences required data, and described inference data is at least
Including one group of data pair with attribute and property value.
Preferably, after building the step of described multiple neural network block forecast model, also include:
Based on predefined feature dictionary, the judgement document in document storehouse is carried out extraction process and obtains the training sample of predetermined theme
Notebook data set；
Based on described training sample data set, training is to obtain described multiple neural network block forecast model；
Wherein it is preferred to, described training sample data set at least includes training sample data, each described training
Sample data all includes that at least five groups are made up of information, demand information, evidence set and court verdict justice court, case respectively
Training data pair；Often organizing described training data to a corresponding data input type, wherein, described justice court is used for determining
Regional information.
Preferably, the logical relation of inference step refers to: in same rule of inference, and the reasoning results of an inference step is this institute
State the final result of rule of inference, or, for the precondition of another inference step；
Described logical relation is for determining the restriction relation between each neural network block, and described restriction relation refers to: same
In forecast model, the output valve of a neural network block is the precondition of the input value of another neural network block, and affects it
Output result.
Preferably, described the provision corresponding with described theme is carried out extraction process and logical process obtains described reasoning rule
Method then includes:
Extract the Feature Words of described provision, form provision characteristic item；
Hierarchical relationship based on provision, determines the logical relation between provision characteristic item；
Rule of inference is built based on the logical relation between provision characteristic item.
Preferably, described based on predefined feature dictionary, judgement document is carried out extraction process and obtains the instruction of predetermined theme
The step practicing sample data sets includes:
At least one judgement document is obtained from document storehouse；
According to predefined feature dictionary, extract instruction corresponding with described predefined Feature Words in each described judgement document
Practice data pair, be correspondingly formed the document characteristic set of each described judgement document；
Document object set is built, the corresponding document feature of each document object based at least one document characteristic set
Set；
Based on default filtering rule, filter the invalid document object in described document object set, obtain effective document pair
As set；
Extract in effective object set at least with the case of described predetermined theme by the corresponding effective object of information, form instruction
Practicing object set, described training object set at least includes a training object；
The data input type of each neural network block based on the forecast model corresponding with predetermined theme, extracts each respectively
Training data pair corresponding with described data input type in training object, to be formed respectively and each described training object phase
Corresponding training sample data；
The number of training of described predetermined theme is built based on the training sample data that each described training object is corresponding
According to set.
Preferably, described default filtering rule includes:
Judge whether the attribute of the training data pair in document characteristic set and/or property value are empty；
If it is empty, then the document object corresponding with described document characteristic set is removed from described document object set；
Judge whether there is case relatedness between at least two document characteristic set；If there is relatedness, then based on literary composition
Court verdict in book characteristic set, determines the document object that the document characteristic set needing to filter out is corresponding；
Wherein, described case relatedness refers to the corresponding same case of respective judgement document of at least two document characteristic set
The different trial programs of part.
Preferably, based on described training sample data set, trained before revising described forecast model, also include:
Attribute based on each training data pair in the predetermined formula training sample data to described predetermined theme and property value are returned
One change processes.
Preferably, based on described training sample data set, training before revising described forecast model, it is right also to include
The redundancy inspection of at least one neural network block described；
Wherein, described redundancy inspection includes: judge to have existed and number of training in current forecast model
The training data with attribute and property value according to is to corresponding neural network block；
If existing, then activation and this described neural network block have other neural network blocks of restriction relation, and just perform
Operate to reasoning；
If not existing, then set up and there is the training data of attribute and property value to corresponding neural network block with described,
And described neural network block is added to current forecast model.
Preferably, it is judged that described execution forward reasoning operation process whether there is contradiction, including:
Judge the output valve of current neural network block whether contradict with the rule of inference in described knowledge reasoning storehouse and/
Or whether the output valve of current neural network block contradicts with the output valve of another neural network block；
If there is contradiction, then described contradiction is set to contradiction point and performs reversely to revise the described nerve that operation correction is current
Network block.
According to another aspect of the present invention, present invention also offers the prognoses system of a kind of lawsuit result, including:
Input block: for obtaining the testing data being currently entered；
Extraction unit: for based on predefined feature dictionary, extract in described testing data with described predefined Feature Words
Corresponding testing data pair, to form testing data characteristic information set；
Subject analysis unit: for based on described testing data characteristic set, determine the applicable theme of described testing data；
Predicting unit, for being suitable for, with described, the prediction mould that theme is corresponding by described testing data characteristic information set input
Type is predicted, and obtains predicting the outcome of described testing data.
Preferably, described prognoses system also includes:
Bar library, for storing the initial data of provision；
Document storehouse, for storing the initial data of existing judgement document；
Knowledge reasoning storehouse, at least one rule of inference that storage is corresponding with described predetermined theme；
Predefined feature dictionary, for storing multiple preset data pair with attribute and property value；Described preset data
To being used for the foundation as the testing data pair extracted in testing data and/or extracting the training data in judgement document to depending on
According to.
Preferably, described prognoses system also includes:
Document acquiring unit, for obtaining judgement document from document storehouse；
Described extraction unit, is additionally operable to described based on predefined feature dictionary, judgement document is carried out extraction process and obtains
The training sample data set of predetermined theme；
Data processing unit, for based on predetermined formula to the attribute of each training data pair in each training sample data and
Attribute and the property value of the testing data pair of property value and/or testing data characteristic information set are normalized；
Training data acquiring unit, for obtaining the training sample data collection after data processing unit processes
Close；
Training unit, for based on training sample data set, operating by forward reasoning and reversely revise operation training
And revise neutral net, so that the actual judgement knot with himself that predicts the outcome that each described training sample data are corresponding
The difference of fruit is in preset range, and the rule of inference with described predetermined theme that makes to predict the outcome matches, thus forms institute
State the neural network prediction model of predetermined theme；
Described predicting unit, is additionally operable to pass through the rational of each neural network block training based on the output of training sample data
Predict the outcome.
Preferably, described prognoses system also includes:
Judging unit, for judging to have existed in current forecast model and having genus in training sample data
The training data of property and property value is to corresponding neural network block；
If existing, then activation and this described neural network block have other neural network blocks of restriction relation, and just perform
Operate to reasoning；
If not existing, then set up and there is the training data of attribute and property value to corresponding neural network block with described,
And described neural network block is added to current forecast model.
For judging whether the process that described execution forward reasoning operates exists contradiction, including:
Judge the output valve of current neural network block whether contradict with the rule of inference in described knowledge reasoning storehouse and/
Or whether the output valve of current neural network block contradicts with the output valve of another neural network block；
If there is contradiction, then described contradiction is set to contradiction point and performs reversely to revise the described nerve that operation correction is current
Network block.
Accompanying drawing explanation
Fig. 1 is the exemplary plot of the multiple neural network block that embodiments of the invention provide；
Fig. 2 is the exemplary plot of the rule of inference of one of them in multiple rule of inference that embodiments of the invention provide.
Detailed description of the invention
It is an object of the invention to provide Forecasting Methodology and the system of a kind of lawsuit result, on the one hand based on China's provision genealogy of law
Feature, from legal provision, the trial rule of technicalization case, builds Analysis of Knowledge Bases Reasoning storehouse, is used for instructing neutral net mould
Type prediction process；On the other hand, based on the big data of real lawsuit, analysis obtains at evidence similar with demand, and evidence proves
Basic case there is similarity, the case trial thinking feature in same region；Under Analysis of Knowledge Bases Reasoning instructs, structure
Building the legal case result Inference Forecast model of multiple neural network block, active client initial conditions is trained prediction may
The trial result obtained, makes up the weak tendency in domestic consumer's general knowledge of laws or analysis ability to a certain extent, strengthens from bodily movement of practising Wushu
The degree of understanding of rule demand.
For making the object, technical solutions and advantages of the present invention of greater clarity, below in conjunction with detailed description of the invention and join
According to accompanying drawing, the present invention is described in more detail.It should be understood that these describe the most exemplary, and it is not intended to limit this
Bright scope.Additionally, in the following description, eliminate the description to known features and technology, to avoid unnecessarily obscuring this
The concept of invention.
The embodiment provides the Forecasting Methodology of a kind of lawsuit result, including:
Obtain testing data；
Based on predefined feature dictionary, extract number to be measured corresponding with described predefined Feature Words in described testing data
According to right, to form testing data characteristic information set；
Based on described testing data characteristic set, determine the applicable theme of described testing data；
Described testing data characteristic information set input is predicted with the described forecast model being suitable for theme corresponding,
To predicting the outcome of described testing data.Predicting the outcome is support possible to demand under the present conditions, is a percentage ratio
Data.In model construction, the expection output result in training sample is to calculate the support of demand based on court verdict
The percentage ratio arrived.For example, plaintiff loses a lawsuit, and that intended output result is 0；Plaintiff's demand is all supported, that intended output
Result is 100%；Plaintiff has 3 demands, and 1 demand is only supported in judgement, and that intended output result is 33.3%.Prediction process
Judge that can current evidence be enough to support demand the most exactly, and calculate supporting degree.For example, the condition of active client
For: deal contract dispute, request the other side pays payment for goods, but the evidence provided only has delivery receipt, then possible prediction knot
Fruit is just 0, and i.e. current lack of evidence, to support plaintiff's demand, only can prove that goods pays the fact, can not proved independent dealing
The establishment of contract, so the request of ignoring.
Described testing data characteristic information set at least includes three be made up of case information, demand information and evidence set
Group testing data pair, often organizes described testing data to a corresponding data type.
As a preferred embodiment, before by described testing data characteristic information set input prediction model, also
Including to the data with attribute and property value in above-mentioned testing data characteristic information set to carrying out data cleansing, data add
Work processes；Described data cleansing includes absent field, mathematical logic, data form and the process of repetition values；Described data
Processing includes data calculating, packet and the process of data conversion.Concrete example and say, extract case in testing data by
Information, regional information determine the current topic of testing data；Wherein, described regional information is based on the justice court in testing data
Determine；
Extract the time data in testing data and form temporal characteristics information；Described time data is used for calculating current time
Whether exceed current case by effective statute of limitation of regulation under information；
Extract the demand information in testing data, Feature Words based on the default demand information corresponding with current topic
Language set is removed the non-feature vocabulary in demand information and is constituted demand information keywords set of words to form demand characteristic information, demand
The size of the key word quantity comprised in characteristic information must be not more than sufficiently large integer T2 set in advance；
Extract the evidence set in testing data, Feature Words based on the default evidence set corresponding with current topic
Language set is extracted the Feature Words in evidence set and is formed evident feature information；
Case is by information, regional information, temporal characteristics information, demand characteristic information and evident feature information the most corresponding one
Input model needed for individual data type, and this data type forecast model that correspondence current topic is corresponding respectively.
Time restriction relation is there is between described temporal characteristics information and demand characteristic information, evidence characteristic information；Described
Example restriction relation is there is between demand characteristic information and evidence characteristic information；Described time-constrain relation refers to constraint element
Time variable value depends on the output valve of nerve net block corresponding to temporal characteristics information, and represents in the meaning of output valve and be in
In the effect lawsuit time limit, retrain follow-up neural network block and can perform normal reasoning process；Otherwise then terminate reasoning；Described demand is special
Reference breath refers to the defeated of the corresponding neural network block of demand characteristic information with the example restriction relation between evidence characteristic information
Whether a certain node going out the value neural network block corresponding with evidence characteristic information exists contradictory relation, if it is present terminate
Reasoning；If it does not exist, then normally perform reasoning process to obtain final the reasoning results；Described blocks of knowledge nkBetween example
Restriction relation refers to blocks of knowledge nk+1Exist according to blocks of knowledge nkThe variable that determines of output valve, and the value of this variable is straight
Connect and affect blocks of knowledge nk+1Output valve.
Before the step that described testing data characteristic information set input multiple neural network forecast model is predicted,
Also include:
Knowledge reasoning storehouse based on the predetermined theme set up, builds described multiple neural network block forecast model.
Described knowledge reasoning storehouse based on the predetermined theme set up, builds the step of described multiple neural network block forecast model
Including:
Set up the knowledge reasoning storehouse of predetermined theme；Wherein, described predetermined theme includes that case is by information；Described knowledge reasoning storehouse
Storage have at least one rule of inference, described rule of inference be by the provision corresponding with described theme is carried out extraction process and
Logical process obtains, and each described rule of inference at least includes multiple inference steps with logical relation；
Based on described knowledge reasoning storehouse, build described multiple neural network block forecast model；
Wherein, each neural network block of described forecast model and a described inference step at least one rule of inference
Corresponding；The corresponding data type of the inference data of each described inference step, each described data type is a nerve net
The data input type of network block；Wherein, the most each described inference step of described inference data makes inferences required data, described in push away
Reason data at least include one group of data pair with attribute and property value.Inference data includes the rule of inference that testing data is corresponding
Data needed for the rule of inference reasoning that data needed for reasoning are corresponding with training sample data.Concrete example illustrates, such as Fig. 1 and
Shown in Fig. 2, the corresponding inference step of little rule of each reasoning in Fig. 2, in the corresponding Fig. 1 of each inference step
One input layer M, and the data needed for each inference step are the data type input type with corresponding input layer M, i.e.
Each input layer M accepts one group of specific attribute and the input of property value.Yet further say, in Fig. 2 " rule 1: whether
Within effective lawsuit phase ", required data just include: standard-terms；occurrence-time；current-time；
Assume the input layer M1 in Fig. 1 for receiving this generic attribute and property value, and occurrence-time=
20160101；Current-time=20160731；Standard-terms=2；Then output layer O1 is then 1, shows that it is in
In effect duration.Contact between each neural network block obtains based on rule of inference；It is assumed that M2The attribute accepted and attribute
Value is " rule 2: whether have written deal contract " desired data, and storehouse rule, only works as O by inference1At=1 before the deadline
When just perform follow-up reasoning, now O1=1, then perform M2Otherwise, then M is performed3.Wherein, it should be noted that, such as Fig. 1 institute
Show, M1、P1、O1Constitute a neural network block, M1Represent input layer, P1Represent intermediate layer, O1For output layer.
After building the step of described multiple neural network block forecast model, also include:
Based on predefined feature dictionary, the judgement document in document storehouse is carried out extraction process and obtains the training sample of predetermined theme
Notebook data set；
Based on described training sample data set, training is to obtain described multiple neural network block forecast model；
Wherein, described training sample data set at least includes training sample data, each described number of training
According to the training all including that at least five groups are made up of information, demand information, evidence set and court verdict justice court, case respectively
Data pair；Often organizing described training data to a corresponding data input type, wherein, described justice court is used for determining that region is believed
Breath.
The logical relation of inference step refers to: in same rule of inference, and the reasoning results of an inference step is this described reasoning
The final result of rule, or, for the precondition of another inference step；
Described logical relation is for determining the restriction relation between each neural network block, and described restriction relation refers to: same
In forecast model, the output valve of a neural network block is the precondition of the input value of another neural network block, and affects it
Output result.
Described the provision corresponding with described theme is carried out extraction process and logical process obtains the side of described rule of inference
Method includes:
Extract the Feature Words of described provision, form provision characteristic item；Specifically, according to provision attribute and property value, right
Provision content carries out participle operation and filter operation, according to the word in filter table, will have little significance provision content recognition, but
Words such as such as ", etc. and or " that frequently occur removes, and generates provision characteristic item.
Hierarchical relationship based on provision, determines the logical relation between provision characteristic item；Described state hierarchical relationship finger literary composition
Between parent relation or sub-level relation, it is also possible to being brotherhood, sub-level provision can be empty；Brother's provision includes its higher level
Non-parent provision, and/or, the non-sub-level provision of subordinate；Including at least the sub-level provision of a non-NULL in parent provision；
Rule of inference is built based on the logical relation between provision characteristic item.
Described based on predefined feature dictionary, judgement document is carried out extraction process and obtains the number of training of predetermined theme
Include according to the step of set:
At least one judgement document is obtained from document storehouse；
According to predefined feature dictionary, extract instruction corresponding with described predefined Feature Words in each described judgement document
Practice data pair, be correspondingly formed the document characteristic set of each described judgement document；As shown in table 1 below, for merchandise building contract
The sample table of one of them document object of the document object set S1 of dispute theme, wherein, the literary composition included by document object
The attribute of book characteristic information i.e. refers to that document Reference Number, case are by information, justice court, trial program, plaintiff/appellant, defendant/upper
Telling people, association Reference Number, demand information, court verdict and acceptance fee, the property value of document characteristic information i.e. refers to each attribute institute
Corresponding respective particular content.It should be noted that only the showing of document object the comprised document characteristic information of this table 1
Example, is not limitation of the invention, and the attribute of document characteristic information and property value are also not limited in table 1 genus that provides
Property and property value.
Table 1
Document object set is built, the corresponding document feature of each document object based at least one document characteristic set
Set；
Based on default filtering rule, filter the invalid document object in described document object set, obtain effective document pair
As set；
Extract in effective object set at least with the case of described predetermined theme by the corresponding effective object of information, form instruction
Practicing object set, described training object set at least includes a training object；
The data input type of each neural network block based on the forecast model corresponding with predetermined theme, extracts each respectively
Training data pair corresponding with described data input type in training object, to be formed respectively and each described training object phase
Corresponding training sample data；
The number of training of described predetermined theme is built based on the training sample data that each described training object is corresponding
According to set.
It should be noted that it is at least corresponding by information with the case of described predetermined theme in extracting effective object set
Effectively object, also includes before forming training object set: the attribute to the data pair of the effective object in effective object set
Data cleansing, data mart modeling is carried out with property value；
Wherein, described data cleansing includes absent field, mathematical logic, data form and the process of repetition values；Institute
State data mart modeling and include data calculating, packet and the process of data conversion.
Concrete data mart modeling and data process and mainly comprise the steps that
The regional information of document object is determined according to the property value of justice court；
The pretreatment of demand information, including: the property value of demand information is carried out participle operation, removes the text of property value
In non-feature vocabulary, formed keyword set；Wherein, according to the feature vocabulary of the demand information under certain default predetermined theme
In set, be not belonging to this feature lexical set is the non-feature vocabulary needing to remove.Especially, the size of keyword set
Sufficiently large integer T2 set in advance must be not more than；
The pretreatment of evidence set, including: according to the feature lexical set of the evidence set under certain default predetermined theme,
Extract key word and form evidence keyword set；
Setting up mapping ruler, the set element e in described evidence keyword set with the predefined category set c preset is
Mapping relations one to one；And differentiate the classification corresponding to described array element e based on mapping ruler；
According to mapping ruler, it is judged that set element ejAffiliated category set, as shown in table 2 below；Work as aijWhen=1, then ej
Belong to ciClass；Otherwise, work as aijWhen=0, then ejIt is not belonging to ciClass；Wherein, array element ejWith predefined category set ciMapping
Relation is for map one to one, i.e. array element ejOnly possible belong to ciClass, and c can not be belonged simultaneously toiClass and ci+jClass.
According to the element classification result obtained by above-mentioned mapping ruler, determine ejClassification, and according to process after classification
Data update array.
Table 2
Acceptance fee and the pretreatment of court verdict, including:
By the process of the property value to acceptance fee, obtain amount of money ratio Cost=undertaken needed for plaintiff/appellant Plf
PPlf；Particularly as follows: judge whether the property value to acceptance fee exists and plaintiff/appellant Plf and/or defendant/appellee D
Identical value, if only existing plaintiff/appellant Plf, then Cost=PPlf=1；If only existing defendant/appellee D, then Cost
=PPlf=0；If there is the value of Plf Yu D simultaneously, and the amount of money should bear mutually is MPlfWith MD, then
According to Cost result, process court verdict Re, obtain plaintiff/appellant's demand degree of support, i.e. Re=(1-
Cost) × 100%=(1-PPlf)×100%。
Described default filtering rule includes:
Judge whether the attribute of the training data pair in document characteristic set and/or property value are empty；
If it is empty, then the document object corresponding with described document characteristic set is removed from described document object set；
Judge whether there is case relatedness between at least two document characteristic set；If there is relatedness, then based on literary composition
Court verdict in book characteristic set, determines the document object that the document characteristic set needing to filter out is corresponding；
Wherein, described case relatedness refers to the corresponding same case of respective judgement document of at least two document characteristic set
The different trial programs of part.As a example by above table 1, table 1 is a document object, and its document Reference Number is (2015) even people's word at end the
No. AAA, association Reference Number is (2015) Lian Dongmin just word the BBBth；" Jiangsu is cancelled according to the court verdict of the document object of table 1
Province Donghai County people's court (2015) Lian Dongmin just word BBB civil judgment ", then by document Reference Number for (2015) Lian Dongmin at the beginning of
The document object that word the BBBth removes from document object set S1.
Based on described training sample data set, trained before revising described forecast model, also include: based on predetermined
Attribute and the property value of each training data pair in the training sample data of described predetermined theme are normalized by formula.
As shown in table 3 below, merely illustrative, not limitation of the invention, the property value of each data pair in described training sample data
It it is a keyword set containing at least one key word；Described predetermined formula isWherein,
J is the jth key word in keyword set, and i represents input layer, βiFor the regulatory factor of described input layer, CijFor key word
KijThe number of times occurred in keyword set；N is the quantity of the key word comprised in keyword set.
Table 3
| Lexical or textual analysis | Mark | Default value | Example | |
| Demand | Appeal | “” | Cancel, the first sentence, judgement, undertake, house purchase fund, one times, reparation | Input |
| Evidence set | Evi | [] | [" contract class ", " Agreement for Sale and Purchase ", " cash voucher "] | Input |
| Court verdict | Re | “” | 100% | Output |
Based on described training sample data set, trained before revising described forecast model, also include to described extremely
The redundancy inspection of a few neural network block；
Wherein, described redundancy inspection includes: judge to have existed and number of training in current forecast model
The training data with attribute and property value according to is to corresponding neural network block；
If existing, then activation and this described neural network block have other neural network blocks of restriction relation, and just perform
Operate to reasoning；
If not existing, then set up and there is the training data of attribute and property value to corresponding neural network block with described,
And described neural network block is added to current forecast model.
Judge whether the process that described execution forward reasoning operates exists contradiction, including:
Judge the output valve of current neural network block whether contradict with the rule of inference in described knowledge reasoning storehouse and/
Or whether the output valve of current neural network block contradicts with the output valve of another neural network block；
If there is contradiction, then described contradiction is set to contradiction point and performs reversely to revise the described nerve that operation correction is current
Network block.
According to another aspect of the present invention, present invention also offers the prognoses system of a kind of lawsuit result, including:
Input block: for obtaining the testing data being currently entered；
Extraction unit: for based on predefined feature dictionary, extract in described testing data with described predefined Feature Words
Corresponding testing data pair, to form testing data characteristic information set；
Subject analysis unit: for based on described testing data characteristic set, determine the applicable theme of described testing data；
Predicting unit, for being suitable for, with described, the prediction mould that theme is corresponding by described testing data characteristic information set input
Type is predicted, and obtains predicting the outcome of described testing data.
Described prognoses system also includes:
Bar library, for storing the initial data of provision；
Document storehouse, for storing the initial data of existing judgement document；
Knowledge reasoning storehouse, at least one rule of inference that storage is corresponding with described predetermined theme；
Predefined feature dictionary, for storing multiple preset data pair with attribute and property value；Described preset data
To being used for the foundation as the testing data pair extracted in testing data and/or extracting the training data in judgement document to depending on
According to.
Described prognoses system also includes: document acquiring unit, for obtaining judgement document from document storehouse；
Described extraction unit, is additionally operable to described based on predefined feature dictionary, judgement document is carried out extraction process and obtains
The training sample data set of predetermined theme；
Data processing unit, for based on predetermined formula to the attribute of each training data pair in each training sample data and
Attribute and the property value of the testing data pair of property value and/or testing data characteristic information set are normalized；
Training data acquiring unit, for obtaining the training sample data collection after data processing unit processes
Close；
Training unit, for based on training sample data set, operating by forward reasoning and reversely revise operation training
And revise neutral net, so that the actual judgement knot with himself that predicts the outcome that each described training sample data are corresponding
The difference of fruit is in preset range, and the rule of inference with described predetermined theme that makes to predict the outcome matches, thus forms institute
State the neural network prediction model of predetermined theme；
Described predicting unit, is additionally operable to pass through the rational of each neural network block training based on the output of training sample data
Predict the outcome.
Described prognoses system also includes: judging unit, for judging to have existed in current forecast model and instruction
Practice the training data with attribute and property value in sample data to corresponding neural network block；If exist, then activate with
This described neural network block has other neural network blocks of restriction relation, and performs forward reasoning operation；If not existing, then build
Vertical have the training data of attribute and property value to corresponding neural network block with described, and is added to by described neural network block
Current forecast model.
For judging whether the process that described execution forward reasoning operates exists contradiction, including:
Judge the output valve of current neural network block whether contradict with the rule of inference in described knowledge reasoning storehouse and/
Or whether the output valve of current neural network block contradicts with the output valve of another neural network block；If there is contradiction, then will
Described contradiction is set to contradiction point and performs reversely to revise the described neural network block that operation correction is current.
It is contemplated that protect Forecasting Methodology and the system of a kind of lawsuit result, on the one hand special based on China's provision genealogy of law
Levy, from legal provision, the trial rule of technicalization case, build Analysis of Knowledge Bases Reasoning storehouse, be used for instructing neural network model
Prediction process；On the other hand, based on the big data of real lawsuit, it is similar with demand that analysis obtains evidence, and the base that evidence proves
This case situation has similarity, the case trial thinking feature in same region；Under Analysis of Knowledge Bases Reasoning instructs, structure is many
The Inference Forecast model of the lawsuit result of neural network block, is trained active client initial conditions predicting and is likely to be obtained
Trial result, makes up domestic consumer's weak tendency in general knowledge of laws or analysis ability to a certain extent, strengthens and needs self law
The degree of understanding asked.
It should be appreciated that the above-mentioned detailed description of the invention of the present invention is used only for exemplary illustration or explains the present invention's
Principle, and be not construed as limiting the invention.Therefore, that is done in the case of without departing from the spirit and scope of the present invention is any
Amendment, equivalent, improvement etc., should be included within the scope of the present invention.Additionally, claims purport of the present invention
Whole within containing the equivalents falling into scope and border or this scope and border change and repair
Change example.
Claims (15)
1. a Forecasting Methodology for lawsuit result, including:
Obtain testing data；
Based on predefined feature dictionary, extract testing data corresponding with described predefined Feature Words in described testing data
Right, to form testing data characteristic information set；Wherein, described testing data characteristic information set at least includes by case by believing
Three groups of testing datas pair that breath, demand information and evidence set are constituted, often organize described testing data to a corresponding data type；
Based on described testing data characteristic set, determine the applicable theme of described testing data；
Described testing data characteristic information set input is predicted with the described forecast model being suitable for theme corresponding, obtains institute
State predicting the outcome of testing data.
Forecasting Methodology the most according to claim 1, it is characterised in that described testing data characteristic information set is inputted
Before the step that multiple neural network forecast model is predicted, also include:
Knowledge reasoning storehouse based on the predetermined theme set up, builds described multiple neural network block forecast model.
Forecasting Methodology the most according to claim 2, it is characterised in that described knowledge reasoning based on the predetermined theme set up
Storehouse, the step building described multiple neural network block forecast model includes:
Set up the knowledge reasoning storehouse of predetermined theme；Wherein, described predetermined theme includes that case is by information and regional information；Described knowledge
Reasoning stock contains at least one rule of inference, and described rule of inference is by extracting the provision corresponding with described theme
Process and logical process obtain, and each described rule of inference at least includes multiple inference steps with logical relation；
Based on described knowledge reasoning storehouse, build described multiple neural network block forecast model；
Wherein, each neural network block of described forecast model and a described inference step pair at least one rule of inference
Should；The corresponding data type of the inference data of each described inference step, each described data type is a neutral net
The data input type of block；Wherein, the most each described inference step of described inference data makes inferences required data, described reasoning
Data at least include one group of data pair with attribute and property value.
Forecasting Methodology the most according to claim 3, it is characterised in that building described multiple neural network block forecast model
After step, also include:
Based on predefined feature dictionary, the judgement document in document storehouse is carried out extraction process and obtains the number of training of predetermined theme
According to set；
Based on described training sample data set, training is to obtain described multiple neural network block forecast model；
Wherein, described training sample data set at least includes training sample data, and each described training sample data are equal
The training data being made up of information, demand information, evidence set and court verdict justice court, case respectively including at least five groups
Right；Often organizing described training data to a corresponding data input type, wherein, described justice court is for domain information definitely.
5. according to the Forecasting Methodology described in Claims 2 or 3, it is characterised in that the logical relation of inference step refers to: in same reasoning
In rule, the reasoning results of an inference step is the final result of this described rule of inference, or, for the preposition bar of another inference step
Part；
Described logical relation is for determining the restriction relation between each neural network block, and described restriction relation refers to: in same prediction
In model, the output valve of a neural network block is the precondition of the input value of another neural network block, and affects its output
Result.
Forecasting Methodology the most according to claim 2, it is characterised in that described the provision corresponding with described theme is carried
Take and process and logical process obtains the method for described rule of inference and includes:
Extract the Feature Words of described provision, form provision characteristic item；
Hierarchical relationship based on provision, determines the logical relation between provision characteristic item；
Rule of inference is built based on the logical relation between provision characteristic item.
Forecasting Methodology the most according to claim 4, it is characterised in that described based on predefined feature dictionary, to judge's literary composition
Book carries out the step of the training sample data set that extraction process obtains predetermined theme and includes:
At least one judgement document is obtained from document storehouse；
According to predefined feature dictionary, extract training number corresponding with described predefined Feature Words in each described judgement document
According to right, it is correspondingly formed the document characteristic set of each described judgement document；
Document object set is built, the corresponding document feature set of each document object based at least one document characteristic set
Close；
Based on default filtering rule, filter the invalid document object in described document object set, obtain effective document object set
Close；
Extract in effective object set at least with the case of described predetermined theme by the corresponding effective object of information, form training right
As set, described training object set at least includes a training object；
The data input type of each neural network block based on the forecast model corresponding with predetermined theme, extracts each training respectively
Training data pair corresponding with described data input type in object, corresponding with each described training object to be formed respectively
Training sample data；
The training sample data collection of described predetermined theme is built based on the training sample data that each described training object is corresponding
Close.
Forecasting Methodology the most according to claim 7, it is characterised in that described default filtering rule includes:
Judge whether the attribute of the training data pair in document characteristic set and/or property value are empty；
If it is empty, then the document object corresponding with described document characteristic set is removed from described document object set；
Judge whether there is case relatedness between at least two document characteristic set；If there is relatedness, then special based on document
Court verdict in collection conjunction, determines the document object that the document characteristic set needing to filter out is corresponding；
Wherein, described case relatedness refers to the corresponding same case of respective judgement document of at least two document characteristic set
Different trial programs.
Forecasting Methodology the most according to claim 4, it is characterised in that based on described training sample data set, training
To revise before described forecast model, also include: each based in the predetermined formula training sample data to described predetermined theme
Attribute and the property value of training data pair are normalized.
Forecasting Methodology the most according to claim 4, it is characterised in that based on described training sample data set, training
Before revising described forecast model, also include the redundancy inspection at least one neural network block described；
Wherein, described redundancy inspection includes: judge current forecast model has existed with in training sample data
The training data with attribute and property value to corresponding neural network block；
If existing, then activate and with this described neural network block there are other neural network blocks of restriction relation, and perform forward and push away
Reason operation；
If not existing, then set up and there is the training data of attribute and property value to corresponding neural network block with described, and will
Described neural network block adds to current forecast model.
11. Forecasting Methodologies according to claim 10, also include, it is judged that whether the process of described execution forward reasoning operation
There is contradiction, including:
Judge that whether the output valve of current neural network block contradicts with the rule of inference in described knowledge reasoning storehouse and/or work as
Whether the output valve of front neural network block contradicts with the output valve of another neural network block；
If there is contradiction, then described contradiction is set to contradiction point and performs reversely to revise the described neutral net that operation correction is current
Block.
The prognoses system of 12. 1 kinds of lawsuit results, including:
Input block: for obtaining the testing data being currently entered；
Extraction unit: for based on predefined feature dictionary, extract in described testing data relative with described predefined Feature Words
The testing data pair answered, to form testing data characteristic information set；
Subject analysis unit: for based on described testing data characteristic set, determine the applicable theme of described testing data；
Predicting unit, for entering described testing data characteristic information set input with the described forecast model being suitable for theme corresponding
Row prediction, obtains predicting the outcome of described testing data.
13. prognoses systems according to claim 12, it is characterised in that described prognoses system also includes:
Bar library, for storing the initial data of provision；
Document storehouse, for storing the initial data of existing judgement document；
Knowledge reasoning storehouse, at least one rule of inference that storage is corresponding with described predetermined theme；
Predefined feature dictionary, for storing multiple preset data pair with attribute and property value；Described preset data to
In as the testing data pair extracted in testing data according to and/or extract the foundation of training data pair in judgement document.
14. prognoses systems according to claim 12, it is characterised in that described prognoses system also includes:
Document acquiring unit, for obtaining judgement document from document storehouse；
Described extraction unit, is additionally operable to described based on predefined feature dictionary, judgement document is carried out extraction process and is made a reservation for
The training sample data set of theme；
Data processing unit, for based on predetermined formula to the attribute of each training data pair in each training sample data and attribute
Attribute and the property value of the testing data pair of value and/or testing data characteristic information set are normalized；
Training data acquiring unit, for obtaining the training sample data set after data processing unit processes；
Training unit, for based on training sample data set, being revised operation training by forward reasoning operation with reverse and repair
Positive neutral net, so that the actual court verdict predicted the outcome with himself corresponding to each described training sample data
Difference is in preset range, and the rule of inference with described predetermined theme that makes to predict the outcome matches, thus is formed described pre-
Determine the neural network prediction model of theme；
Described predicting unit, is additionally operable to predict through the rational of each neural network block training based on the output of training sample data
Result.
15. prognoses systems according to claim 12, it is characterised in that described prognoses system also includes:
Judging unit, for judge current forecast model has existed with in training sample data have attribute and
The training data of property value is to corresponding neural network block；
If existing, then activate and with this described neural network block there are other neural network blocks of restriction relation, and perform forward and push away
Reason operation；
If not existing, then set up and there is the training data of attribute and property value to corresponding neural network block with described, and will
Described neural network block adds to current forecast model；
For judging whether the process that described execution forward reasoning operates exists contradiction, including:
Judge that whether the output valve of current neural network block contradicts with the rule of inference in described knowledge reasoning storehouse and/or work as
Whether the output valve of front neural network block contradicts with the output valve of another neural network block；
If there is contradiction, then described contradiction is set to contradiction point and performs reversely to revise the described neutral net that operation correction is current
Block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610647687.7A CN106296495A (en) | 2016-08-09 | 2016-08-09 | The Forecasting Methodology of a kind of lawsuit result and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610647687.7A CN106296495A (en) | 2016-08-09 | 2016-08-09 | The Forecasting Methodology of a kind of lawsuit result and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106296495A true CN106296495A (en) | 2017-01-04 |
Family
ID=57667191
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610647687.7A Pending CN106296495A (en) | 2016-08-09 | 2016-08-09 | The Forecasting Methodology of a kind of lawsuit result and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106296495A (en) |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107562938A (en) * | 2017-09-21 | 2018-01-09 | 重庆工商大学 | A kind of law court intelligently tries method |
| CN107632968A (en) * | 2017-05-22 | 2018-01-26 | 南京大学 | A kind of construction method of chain of evidence relational model towards judgement document |
| CN107894981A (en) * | 2017-12-13 | 2018-04-10 | 武汉烽火普天信息技术有限公司 | A kind of automatic abstracting method of case semantic feature |
| CN107918921A (en) * | 2017-11-21 | 2018-04-17 | 南京擎盾信息科技有限公司 | Criminal case court verdict measure and system |
| CN108133436A (en) * | 2017-11-23 | 2018-06-08 | 科大讯飞股份有限公司 | Automatic method and system of deciding a case |
| 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 |
| CN108596360A (en) * | 2018-03-16 | 2018-09-28 | 北京中科闻歌科技股份有限公司 | A kind of judgement prediction technique and system based on machine learning |
| CN109308355A (en) * | 2018-09-17 | 2019-02-05 | 清华大学 | Legal decision prediction of result method and device |
| CN109359175A (en) * | 2018-09-07 | 2019-02-19 | 平安科技（深圳）有限公司 | Electronic device, the method for lawsuit data processing and storage medium |
| CN109409625A (en) * | 2018-08-16 | 2019-03-01 | 上海绿狮智能信息科技股份有限公司 | The method that a kind of pair of legal affairs carries out quantitative evaluation |
| CN109410096A (en) * | 2018-10-18 | 2019-03-01 | 上海右上角文化传媒有限公司 | Method and equipment for providing litigation strategy information of infringing case |
| CN110020134A (en) * | 2017-11-09 | 2019-07-16 | 北京国双科技有限公司 | A kind of method for pushing, system, storage medium and the processor of knowledge services information |
| CN110222145A (en) * | 2019-04-28 | 2019-09-10 | 杭州实在智能科技有限公司 | A kind of intelligence law appraisal procedure and system |
| CN110222866A (en) * | 2019-04-28 | 2019-09-10 | 杭州实在智能科技有限公司 | In conjunction with the intelligent civil case forecasting system and method for colloquial style description and question and answer |
| CN110232447A (en) * | 2019-04-28 | 2019-09-13 | 杭州实在智能科技有限公司 | Legal case depth reasoning method |
| CN110288495A (en) * | 2019-06-25 | 2019-09-27 | 北京市律典通科技有限公司 | Case statute of limitation intelligence checking method and device |
| CN110472011A (en) * | 2019-07-19 | 2019-11-19 | 平安科技（深圳）有限公司 | A kind of cost of litigation prediction technique, device and terminal device |
| CN110674840A (en) * | 2019-08-22 | 2020-01-10 | 中国司法大数据研究院有限公司 | Multi-party evidence association model construction method based on Bayesian network and evidence chain extraction method and device |
| CN110675023A (en) * | 2019-08-22 | 2020-01-10 | 中国司法大数据研究院有限公司 | Litigation request rationality prediction model training method based on neural network, and litigation request rationality prediction method and device based on neural network |
| CN110738032A (en) * | 2018-07-03 | 2020-01-31 | 北京国双科技有限公司 | method and device for generating judgment sections of official paperwork institute |
| CN110770781A (en) * | 2017-06-13 | 2020-02-07 | 本田技研工业株式会社 | Legal information processing system, method and program |
| CN110837562A (en) * | 2018-08-17 | 2020-02-25 | 阿里巴巴集团控股有限公司 | Case processing method, device and system |
| CN110909914A (en) * | 2019-10-12 | 2020-03-24 | 中国平安财产保险股份有限公司 | Litigation success rate prediction method, litigation success rate prediction device, computer device, and storage medium |
| CN110968688A (en) * | 2018-09-30 | 2020-04-07 | 北京国双科技有限公司 | Judicial data processing method and system |
| CN111127183A (en) * | 2018-10-31 | 2020-05-08 | 重庆小雨点小额贷款有限公司 | Data processing method, device, server and computer readable storage medium |
| CN111382769A (en) * | 2018-12-29 | 2020-07-07 | 阿里巴巴集团控股有限公司 | Information processing method, device and system |
| CN112825174A (en) * | 2019-11-15 | 2021-05-21 | 阿里云计算有限公司 | Litigation prediction method, device, system and computer storage medium |
| CN115796285A (en) * | 2023-02-13 | 2023-03-14 | 上海百事通法务信息技术有限公司浙江分公司 | Litigation case prejudging method and device based on engineering model and electronic equipment |
| CN116523036A (en) * | 2023-04-13 | 2023-08-01 | 北京鹅厂科技有限公司 | Method and device for making intermediate decisions and suggesting by multiple persons through large model |
-
2016
- 2016-08-09 CN CN201610647687.7A patent/CN106296495A/en active Pending
Cited By (43)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN107632968A (en) * | 2017-05-22 | 2018-01-26 | 南京大学 | A kind of construction method of chain of evidence relational model towards judgement document |
| CN110770781A (en) * | 2017-06-13 | 2020-02-07 | 本田技研工业株式会社 | Legal information processing system, method and program |
| US11715045B2 (en) | 2017-06-13 | 2023-08-01 | Honda Motor Co, Ltd. | Legal information processing system, method, and non-transitory computer-readable storage medium storing program |
| CN107562938A (en) * | 2017-09-21 | 2018-01-09 | 重庆工商大学 | A kind of law court intelligently tries method |
| CN110020134B (en) * | 2017-11-09 | 2021-08-13 | 北京国双科技有限公司 | Knowledge service information pushing method and system, storage medium and processor |
| CN110020134A (en) * | 2017-11-09 | 2019-07-16 | 北京国双科技有限公司 | A kind of method for pushing, system, storage medium and the processor of knowledge services information |
| CN107918921A (en) * | 2017-11-21 | 2018-04-17 | 南京擎盾信息科技有限公司 | Criminal case court verdict measure and system |
| CN107918921B (en) * | 2017-11-21 | 2021-10-08 | 南京擎盾信息科技有限公司 | Criminal case judgment result measuring method and system |
| CN108133436A (en) * | 2017-11-23 | 2018-06-08 | 科大讯飞股份有限公司 | Automatic method and system of deciding a case |
| CN107894981A (en) * | 2017-12-13 | 2018-04-10 | 武汉烽火普天信息技术有限公司 | A kind of automatic abstracting method of case semantic feature |
| 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 |
| CN108596360A (en) * | 2018-03-16 | 2018-09-28 | 北京中科闻歌科技股份有限公司 | A kind of judgement prediction technique and system based on machine learning |
| CN110738032B (en) * | 2018-07-03 | 2024-02-13 | 北京国双科技有限公司 | Method and device for generating judge paperwork thinking section |
| CN110738032A (en) * | 2018-07-03 | 2020-01-31 | 北京国双科技有限公司 | method and device for generating judgment sections of official paperwork institute |
| CN109409625A (en) * | 2018-08-16 | 2019-03-01 | 上海绿狮智能信息科技股份有限公司 | The method that a kind of pair of legal affairs carries out quantitative evaluation |
| CN109409625B (en) * | 2018-08-16 | 2021-09-14 | 上海绿狮智能信息科技股份有限公司 | Method for quantitatively evaluating legal affairs |
| CN110837562A (en) * | 2018-08-17 | 2020-02-25 | 阿里巴巴集团控股有限公司 | Case processing method, device and system |
| CN110837562B (en) * | 2018-08-17 | 2023-05-02 | 阿里巴巴集团控股有限公司 | Case processing method, device and system |
| CN109359175B (en) * | 2018-09-07 | 2023-04-07 | 平安科技（深圳）有限公司 | Electronic device, litigation data processing method, and storage medium |
| CN109359175A (en) * | 2018-09-07 | 2019-02-19 | 平安科技（深圳）有限公司 | Electronic device, the method for lawsuit data processing and storage medium |
| CN109308355A (en) * | 2018-09-17 | 2019-02-05 | 清华大学 | Legal decision prediction of result method and device |
| CN109308355B (en) * | 2018-09-17 | 2020-03-13 | 清华大学 | Legal judgment result prediction method and device |
| CN110968688A (en) * | 2018-09-30 | 2020-04-07 | 北京国双科技有限公司 | Judicial data processing method and system |
| CN109410096A (en) * | 2018-10-18 | 2019-03-01 | 上海右上角文化传媒有限公司 | Method and equipment for providing litigation strategy information of infringing case |
| CN109410096B (en) * | 2018-10-18 | 2024-02-06 | 上海右云信息技术有限公司 | Method and equipment for providing litigation strategy information of infringing cases |
| CN111127183A (en) * | 2018-10-31 | 2020-05-08 | 重庆小雨点小额贷款有限公司 | Data processing method, device, server and computer readable storage medium |
| CN111127183B (en) * | 2018-10-31 | 2023-10-17 | 重庆小雨点小额贷款有限公司 | Data processing method, device, server and computer readable storage medium |
| CN111382769A (en) * | 2018-12-29 | 2020-07-07 | 阿里巴巴集团控股有限公司 | Information processing method, device and system |
| CN111382769B (en) * | 2018-12-29 | 2023-09-22 | 阿里巴巴集团控股有限公司 | Information processing method, device and system |
| CN110222145A (en) * | 2019-04-28 | 2019-09-10 | 杭州实在智能科技有限公司 | A kind of intelligence law appraisal procedure and system |
| CN110222866A (en) * | 2019-04-28 | 2019-09-10 | 杭州实在智能科技有限公司 | In conjunction with the intelligent civil case forecasting system and method for colloquial style description and question and answer |
| CN110232447A (en) * | 2019-04-28 | 2019-09-13 | 杭州实在智能科技有限公司 | Legal case depth reasoning method |
| CN110288495A (en) * | 2019-06-25 | 2019-09-27 | 北京市律典通科技有限公司 | Case statute of limitation intelligence checking method and device |
| CN110472011A (en) * | 2019-07-19 | 2019-11-19 | 平安科技（深圳）有限公司 | A kind of cost of litigation prediction technique, device and terminal device |
| CN110472011B (en) * | 2019-07-19 | 2023-07-14 | 平安科技（深圳）有限公司 | Litigation cost prediction method and device and terminal equipment |
| CN110675023B (en) * | 2019-08-22 | 2022-03-25 | 中国司法大数据研究院有限公司 | Litigation request rationality prediction model training method based on neural network, and litigation request rationality prediction method and device based on neural network |
| CN110674840A (en) * | 2019-08-22 | 2020-01-10 | 中国司法大数据研究院有限公司 | Multi-party evidence association model construction method based on Bayesian network and evidence chain extraction method and device |
| CN110675023A (en) * | 2019-08-22 | 2020-01-10 | 中国司法大数据研究院有限公司 | Litigation request rationality prediction model training method based on neural network, and litigation request rationality prediction method and device based on neural network |
| CN110909914A (en) * | 2019-10-12 | 2020-03-24 | 中国平安财产保险股份有限公司 | Litigation success rate prediction method, litigation success rate prediction device, computer device, and storage medium |
| CN112825174A (en) * | 2019-11-15 | 2021-05-21 | 阿里云计算有限公司 | Litigation prediction method, device, system and computer storage medium |
| CN115796285A (en) * | 2023-02-13 | 2023-03-14 | 上海百事通法务信息技术有限公司浙江分公司 | Litigation case prejudging method and device based on engineering model and electronic equipment |
| CN116523036A (en) * | 2023-04-13 | 2023-08-01 | 北京鹅厂科技有限公司 | Method and device for making intermediate decisions and suggesting by multiple persons through large model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106296495A (en) | The Forecasting Methodology of a kind of lawsuit result and system | |
| US11062094B2 (en) | Systems and methods for automatically detecting sentiments and assigning and analyzing quantitate values to the sentiments expressed in text | |
| Li et al. | Contextual recommendation based on text mining | |
| TW201537367A (en) | Providing search recommendation | |
| Khotimah et al. | Sentiment analysis of hotel aspect using probabilistic latent semantic analysis, word embedding and LSTM | |
| CN105741077A (en) | Occupational information providing method and system | |
| Gievska et al. | A hybrid approach for emotion detection in support of affective interaction | |
| Ayoub et al. | Analyzing customer needs of product ecosystems using online product reviews | |
| Nedeljković et al. | Building a construction project key-phrase network from unstructured text documents | |
| Romli et al. | Eco-design case-based reasoning tool: The integration of ecological quality function deployment and case-based reasoning methods for supporting sustainable product design | |
| Kim et al. | Accurate and prompt answering framework based on customer reviews and question-answer pairs | |
| Gozuacik et al. | Technological forecasting based on estimation of word embedding matrix using LSTM networks | |
| Cabrio et al. | A support framework for argumentative discussions management in the web | |
| Su et al. | An improved BERT method for the evolution of network public opinion of major infectious diseases: Case Study of COVID-19 | |
| US20140272842A1 (en) | Assessing cognitive ability | |
| Angamuthu et al. | Integrating multi-criteria decision-making with hybrid deep learning for sentiment analysis in recommender systems | |
| US20230325424A1 (en) | Systems and methods for generating codes and code books based using cosine proximity | |
| Rus et al. | Measuring hotel service quality in borobudur temple using opinion mining | |
| Subowo et al. | Twitter Data as Decision Tree Parameter for Analysis of Tourism Potential Policies | |
| Sboev et al. | A comparison of Data Driven models of solving the task of gender identification of author in Russian language texts for cases without and with the gender deception | |
| Soni et al. | Comparative analysis of rotten tomatoes movie reviews using sentiment analysis | |
| Muchmore | Marketing Authorization at the FDA: Paradigms and Alternatives | |
| Chakrabarti | Artificial conversations for chatter bots using knowledge representation, learning, and pragmatics | |
| Lee et al. | Building place-specific sentiment Lexicon | |
| Wambsganss et al. | Using Deep Learning for Extracting User-Generated Knowledge from Web Communities. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170104 |