CN109992664A - Mark classification method, device, computer equipment and the storage medium of central issue - Google Patents
Mark classification method, device, computer equipment and the storage medium of central issue Download PDFInfo
- Publication number
- CN109992664A CN109992664A CN201910185041.5A CN201910185041A CN109992664A CN 109992664 A CN109992664 A CN 109992664A CN 201910185041 A CN201910185041 A CN 201910185041A CN 109992664 A CN109992664 A CN 109992664A
- Authority
- CN
- China
- Prior art keywords
- focus
- central issue
- type
- model
- focus type
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000000605 extraction Methods 0.000 claims abstract description 49
- 230000015654 memory Effects 0.000 claims description 20
- 238000004590 computer program Methods 0.000 claims description 16
- 238000012545 processing Methods 0.000 claims description 16
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 6
- 229940074869 marquis Drugs 0.000 claims description 4
- VBUNOIXRZNJNAD-UHFFFAOYSA-N ponazuril Chemical compound CC1=CC(N2C(N(C)C(=O)NC2=O)=O)=CC=C1OC1=CC=C(S(=O)(=O)C(F)(F)F)C=C1 VBUNOIXRZNJNAD-UHFFFAOYSA-N 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 abstract description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000003247 decreasing effect Effects 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 7
- 238000005457 optimization Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 239000000571 coke Substances 0.000 description 6
- 238000012795 verification Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004939 coking Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000013518 transcription Methods 0.000 description 1
- 230000035897 transcription Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/247—Thesauruses; Synonyms
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Technology Law (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
This application involves field of artificial intelligence, are applied to smart city industry, more particularly to mark classification method, device, computer equipment and the storage medium of a kind of central issue.Method in one embodiment includes: to obtain focus locating rule and judgement document's sample, focal position mark is carried out to judgement document's sample according to focus locating rule, obtain judgement document's sample with mark, extract the central issue in judgement document's sample with mark, and obtain the list of types of disaggregated model and central issue, according to disaggregated model and the list of types of central issue, classify to the central issue of extraction, the automatic classification results for the central issue extracted.Auto-focusing point may be implemented in this way and be labeled classification, the workload of professional mark personnel can be greatly decreased, classification is labeled by model focus point, the mark that can effectively shorten is classified the period, and the working efficiency of mark classification is improved.
Description
Technical field
This application involves field of artificial intelligence, more particularly to a kind of mark classification method of central issue, device,
Computer equipment and storage medium.
Background technique
Judgement document is for recording people's court's hearing process and as a result, being carrier and the people of lawsuit action result
Law court is determining and distributes the voucher of party's substantive right obligation.
Traditional, it needs manually to carry out positioning classification to the central issue in judgement document, due to strongly professional, training
It is difficult, the features such as mark amount is big, often more difficult to be completed in a short time task, lead to mark classification work low efficiency.
Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of mark of central issue that can be improved working efficiency
Infuse classification method, device, computer equipment and storage medium.
A kind of mark classification method of central issue, which comprises
Focus locating rule and judgement document's sample are obtained, the focus locating rule is used to position the coke of judgement document
Point position；
Focal position mark is carried out to judgement document's sample according to the focus locating rule, obtains the sanction with mark
Sentence document sample；
The central issue in judgement document's sample of band mark is extracted, and obtains disaggregated model and central issue
List of types；
According to the disaggregated model and the list of types of the central issue, classify to the central issue of extraction,
Obtain the automatic classification results of the central issue of the extraction.
In one embodiment, the disaggregated model includes rule model and similarity model, described according to point
The list of types of class model and the central issue classifies to the central issue of extraction, obtains the dispute of the extraction
The automatic classification results of focus, comprising:
According to the central issue of the list of types of the central issue, the rule model and extraction, the rule are obtained
The then corresponding focus type candidate collection of model；
According to the central issue of the list of types of the central issue, the similarity model and extraction, obtain described
The corresponding focus type candidate collection of similarity model；
According to the corresponding focus type candidate collection of the rule model and the corresponding focus type of the similarity model
Candidate collection, the corresponding focus type of the central issue extracted.
In one embodiment, described according to the list of types of the central issue, the rule model and extraction
Central issue obtains the corresponding focus type candidate collection of the rule model, comprising:
It is removed redundancy word and the processing of shared word by central issue of the rule model to extraction, and is carried out synonymous
Word conversion process；
According to the list of types of the central issue, central issue all types of in processed central issue is counted, it is raw
Coking vertex type keyword dictionary；
The focus sentence in judgement document's sample is obtained, redundancy word is removed to the focus sentence and synonym turns
Change processing；
The focus descriptor in processed focus sentence is obtained, the focus descriptor is successively closed with the focus type
Keyword dictionary is compared, and is obtained each focus descriptor and is belonged to the general of central issue type in the focus type keyword dictionary
Rate value；
The focus type keyword dictionary is traversed, probability value is chosen and is greater than the central issue type of preset threshold as rule
The then focus type candidate collection of model.
In one embodiment, described according to the list of types of the central issue, the similarity model and extraction
Central issue, obtain the corresponding focus type candidate collection of the similarity model, comprising:
Redundancy word is removed to the central issue extracted by similarity model and shared word is handled, and is carried out synonymous
Word conversion process；
According to the list of types of the central issue, central issue all types of in processed central issue is counted, it is raw
At focus keyword；
Each central issue in the focus keyword is converted into central issue term vector, with central issue term vector
Mean value obtains focus type sentence vector as central issue sentence vector；
The focus sentence in judgement document's sample is obtained, the focus sentence is converted into focus and describes sentence vector；
The focus is described into sentence vector and traverses the focus type sentence vector, focus type is obtained by similarity calculation
Distance value, selected distance value meets focus type candidate collection of the focus type as similarity model of preset condition.
In one embodiment, described according to the corresponding focus type candidate collection of the rule model and the similarity
The corresponding focus type candidate collection of model, the corresponding focus type of the central issue extracted, comprising:
Seek the corresponding focus type candidate collection of the rule model and the corresponding focus type of the similarity model
The intersection of candidate collection；
When the number for handing over focalization type is 1, using the focus type in the intersection as the dispute extracted
The corresponding focus type of focus.
In one embodiment, described to seek the corresponding focus type candidate collection of the rule model and the similarity
After the intersection of the corresponding focus type candidate collection of model, further includes:
When the number for handing over focalization type is greater than 1, the probability value of each focus type and corresponding is obtained
Distance value；
The probability value of each focus type and the product of respective distances value inverse are calculated, by the maximum focus class of product
Type is as the corresponding focus type of central issue extracted.
In one embodiment, described to seek the corresponding focus type candidate collection of the rule model and the similarity
After the intersection of the corresponding focus type candidate collection of model, further includes:
When the number for handing over focalization type is 0, the probability value of each focus type is obtained；
Using the maximum focus type of probability value as the corresponding focus type of central issue extracted.
A kind of mark sorter of central issue, described device include:
Rule acquisition module, for obtaining focus locating rule and judgement document's sample, the focus locating rule is used
In the focal position of positioning judgement document；
Sample labeling module, for carrying out focal position mark to judgement document's sample according to the focus locating rule
Note obtains judgement document's sample with mark；
Focus extraction module, the central issue in judgement document's sample for extracting the band mark, and obtain classification
The list of types of model and central issue；
Classification of focal spot module, for the list of types according to the disaggregated model and the central issue, to extraction
Central issue is classified, and the automatic classification results of the central issue of the extraction are obtained.
A kind of computer equipment, including memory and processor, the memory are stored with computer program, the processing
Device performs the steps of when executing the computer program
Focus locating rule and judgement document's sample are obtained, the focus locating rule is used to position the coke of judgement document
Point position；
Focal position mark is carried out to judgement document's sample according to the focus locating rule, obtains the sanction with mark
Sentence document sample；
The central issue in judgement document's sample of band mark is extracted, and obtains disaggregated model and central issue
List of types；
According to the disaggregated model and the list of types of the central issue, classify to the central issue of extraction,
Obtain the automatic classification results of the central issue of the extraction.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
It is performed the steps of when row
Focus locating rule and judgement document's sample are obtained, the focus locating rule is used to position the coke of judgement document
Point position；
Focal position mark is carried out to judgement document's sample according to the focus locating rule, obtains the sanction with mark
Sentence document sample；
The central issue in judgement document's sample of band mark is extracted, and obtains disaggregated model and central issue
List of types；
According to the disaggregated model and the list of types of the central issue, classify to the central issue of extraction,
Obtain the automatic classification results of the central issue of the extraction.
Mark classification method, device, computer equipment and the storage medium of above-mentioned central issue, by obtaining focus positioning
Rule and judgement document's sample carry out focal position mark to judgement document's sample according to focus locating rule, obtain band mark
Judgement document's sample of note extracts the central issue in judgement document's sample with mark, and obtains disaggregated model and dispute
The list of types of focus is classified to the central issue of extraction, is obtained according to disaggregated model and the list of types of central issue
To the automatic classification results of the central issue of extraction, auto-focusing point may be implemented in this way and be labeled classification, can substantially subtract
The workload of few profession mark personnel, is labeled classification by model focus point, and the mark that can effectively shorten is classified the period, mentions
The working efficiency of height mark classification.
Detailed description of the invention
Fig. 1 is the flow diagram of the mark classification method of central issue in one embodiment；
Fig. 2 is the flow diagram of classification of focal spot step in one embodiment；
Fig. 3 is the flow diagram of rule model focus type Candidate Set obtaining step in one embodiment；
Fig. 4 is the flow diagram of similarity model focus type Candidate Set obtaining step in one embodiment；
Fig. 5 is the flow diagram of focus type determination step in one embodiment；
Fig. 6 is the structural block diagram of the mark sorter of central issue in one embodiment；
Fig. 7 is the internal structure chart of computer equipment in one embodiment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
In one embodiment, as shown in Figure 1, providing a kind of mark classification method of central issue, including following step
It is rapid:
Step 102, focus locating rule and judgement document's sample are obtained, focus locating rule is for positioning judgement document
Focal position.
Focus locating rule refers to the rule for positioning focal position in judgement document, and focus locating rule can pass through
Expert provides.Focus locating rule specifically can be " ' focus ' printed words can be had before central issue content ", " central issue appearance
At judgement document end ", " having law court's viewpoint behind central issue to show to support or do not support the focus " etc..Focus positioning rule
Independence is not needed then, also need not be accurate, fuzzy performance enhances generalization ability.
Judgement document's sample refer to central issue carry out automatic marking classification based training when judgement document, such as ten thousand parts with
Judgement document's training sample of upper magnitude.Document central issue in judgement document's training sample was not both positioned, and was not also had
Labeled classification.
Step 104, focal position mark is carried out to judgement document's sample according to focus locating rule, obtains the sanction with mark
Sentence document sample.
It is the label equation described using program language by focus locating rule transcription, passes through the generation in third party's open source library
Model obtains judgement document's sample with mark, generates model for investigating different label equations, by generating model output
The weight and prediction result of label equation.For example, being examined by the Generative Model in third party's open source library Snorkel
Examine different label equations.The character string of the methods of regular expression, Keywords matching processing input can be used in label equation
And return the result, such as a sentence in input judgement document, one two classification value of corresponding output, 1 indicates that the sentence is dispute
Focus describes sentence, and it is that central issue describes sentence that 0, which indicates the sentence not,.
Step 106, the central issue in judgement document's sample with mark is extracted, and obtains disaggregated model and dispute coke
The list of types of point.
The list of types of central issue refers to the inventory including various central issue types, and the list of types of central issue can
To be provided by expert, for example judgement document's case is determined first by, i.e. document classification, then to provide the dispute occurred under such document burnt
The list of types of point.List of types specifically can be with phrase form, such as " XXX identification ", " XXX identification " etc., specifically can be with
It is " legal relation recognizes (class) ", " authenticity identification (class) " etc..Disaggregated model refers to for focus description to be mapped to correspondence
Focus type model.
In one embodiment, the judgement document with mark can be extracted by the deep neural network model trained
Central issue in sample.For example, the embeding layer of deep neural network model switchs to feature for sentence is inputted using Transformer
Then vector passes through the coding of two-way LSTM (Long Short-Term Memory, shot and long term memory network), finally uses
CRF algorithm (Conditional Random Field, condition random field algorithm) is judged.Deep neural network mould
Type has robustness to label noise, thus be applicable in by third party increase income library Snorkel it is unsupervised predict obtain band mark
Judgement document's sample.By being trained to deep neural network model, can extract in judgement document's sample with mark
Central issue.
Step 108, according to disaggregated model and the list of types of central issue, classify to the central issue of extraction,
The automatic classification results for the central issue extracted.
According to the focus type in the list of types of central issue, divided by central issue of the disaggregated model to extraction
Class exports the corresponding focus type of central issue of extraction.Automatic classification results refer to that the central issue automatically generated is corresponding
Focus type.
The mark classification method of above-mentioned central issue, by obtaining focus locating rule and judgement document's sample, according to
Focus locating rule carries out focal position mark to judgement document's sample, obtains judgement document's sample with mark, extracts band mark
Central issue in judgement document's sample of note, and the list of types of disaggregated model and central issue is obtained, according to classification mould
The list of types of type and central issue classifies to the central issue of extraction, and automatic point of the central issue extracted
The workload of professional mark personnel can be greatly decreased as a result, auto-focusing point may be implemented be labeled classification in this way in class, logical
It crosses model focus point and is labeled classification, the mark that can effectively shorten is classified the period, and the working efficiency of mark classification is improved.
In one embodiment, disaggregated model includes rule model and similarity model, as shown in Fig. 2, according to classification
The list of types of model and central issue classifies to the central issue of extraction, the central issue extracted it is automatic
Classification results, comprising: step 202, according to the list of types of central issue, rule model and the central issue of extraction, obtain
The corresponding focus type candidate collection of rule model；Step 204, it according to the list of types of central issue, similarity model and mentions
The central issue taken obtains the corresponding focus type candidate collection of similarity model；Step 206, according to the corresponding coke of rule model
Vertex type candidate collection and the corresponding focus type candidate collection of similarity model, the corresponding focus class of the central issue extracted
Type.Rule model refers to based on rule is preset, and obtains the model that central issue belongs to the probability value of certain class focus type, phase
Refer to the model that central issue Yu each focus type distance value are obtained by similarity calculation like degree model.
In one embodiment, as shown in figure 3, striving according to the list of types of central issue, rule model and extraction
Focus is discussed, the corresponding focus type candidate collection of rule model is obtained, comprising: step 302, the dispute by rule model to extraction
Focus is removed redundancy word and the processing of shared word, and carries out synonym conversion process；Step 304, according to the class of central issue
Type list counts central issue all types of in processed central issue, generates focus type keyword dictionary；Step 306,
The focus sentence in judgement document's sample is obtained, focus point sentence is removed redundancy word and synonym conversion process；Step 308,
The focus descriptor in processed focus sentence is obtained, focus descriptor is successively compared with focus type keyword dictionary
Compared with obtaining the probability value that each focus descriptor belongs to central issue type in focus type keyword dictionary；Step 310, it traverses
Focus type keyword dictionary chooses probability value and is greater than focus type of the central issue type of preset threshold as rule model
Candidate collection.
Rule model removes redundancy word and shared word first, and makees synonym conversion, then counts each focus type packet
The keyword contained generates focus type keyword dictionary.For using legal relation identification as focus type, such focus type
Including keyword can be debtor-creditor relationship, investment relation, cooperative relationship, unjustified enrichment, debt, credits etc., different focal point class
The corresponding keyword of type is preferably no or rare intersection.When input focus sentence, focus point sentence is removed redundancy word and same
Adopted word conversion process.Successively compare listed candidate word in focus descriptor in focus sentence and focus type keyword dictionary, two
The ratio between the identical word number of person and certain focus type keyword dictionary number, as the focus descriptor is predicted to be the focus type
Probability value.Focus type keyword dictionary is traversed, focus class of the maximum focus type of k probability value as rule model before taking
Type candidate collection.For example, the focus descriptor in focus sentence includes a1 and b2, listed candidate word packet in focus type keyword dictionary
Focus type A, B and C are included, focus type A includes keyword a1, a2 and a3, and focus type B includes keyword b1 and b2, focus
Type C includes keyword a1, a2, b1 and b2.Focus descriptor a1 word number identical with candidate word listed in focus type A is 1,
Focus type A keyword dictionary number is 3, then the probability value that focus descriptor a1 is predicted to be focus type A is 1/3.It is burnt
Point descriptor a1 word number identical with candidate word listed in focus type B is 0, and focus type B keyword dictionary number is 2, then
The probability value that focus descriptor a1 is predicted to be focus type B is 0.Focus descriptor a1 and listed candidate in focus Type C
The identical word number of word is 1, and focus Type C keyword dictionary number is 4, then focus descriptor a1 is predicted to be focus type A's
Probability value is 1/4.Focus descriptor b1 word number identical with candidate word listed in focus type A is 0, and focus type A is crucial
Word dictionary number is 3, then the probability value that focus descriptor b1 is predicted to be focus type A is 0.Focus descriptor b1 and focus
The identical word number of listed candidate word is 1 in type B, and focus type B keyword dictionary number is 2, then focus descriptor b1 is pre-
Surveying as the probability value of focus type B is 1/2.Focus descriptor b1 word number identical with candidate word listed in focus Type C is
1, focus Type C keyword dictionary number is 4, then the probability value that focus descriptor b1 is predicted to be focus Type C is 1/4.
Assuming that k=2, chooses focus type candidate collection of the preceding maximum focus type of 2 probability values as rule model, i.e. the focus sentence
The focus type candidate collection of corresponding rule model includes focus type A and B.
In one embodiment, as shown in figure 4, according to the list of types of central issue, similarity model and extraction
Central issue obtains the corresponding focus type candidate collection of similarity model, comprising: step 402, by similarity model to extraction
Central issue out is removed redundancy word and the processing of shared word, and carries out synonym conversion process；Step 404, according to dispute
The list of types of focus counts central issue all types of in processed central issue, generates focus keyword；Step 406,
Each central issue in focus keyword is converted into central issue term vector, using the mean value of central issue term vector as dispute
Focus sentence vector obtains focus type sentence vector；Step 408, the focus sentence in judgement document's sample is obtained, focus sentence is converted
Sentence vector is described for focus；Step 410, focus is described into sentence vector traversal focus type sentence vector, is obtained by similarity calculation
To the distance value of focus type, selected distance value meets focus type marquis of the focus type as similarity model of preset condition
Selected works.
Similarity model removes redundancy word and shared word first, and makees synonym conversion, generates focus keyword.Then make
Focus keyword is converted into term vector with skip-gram algorithm, then using the mean value of the type central issue term vector as should
The sentence vector of focus type.When input focus sentence, handled by same steps, obtain focus sentence focus describe sentence to
Amount.Focus describes sentence vector traversal focus type sentence vector, carries out similarity calculation two-by-two and obtains distance value, specifically can be Europe
Formula distance, focus type candidate collection of the smallest focus type of k distance value as similarity model before taking.
In one embodiment, as shown in figure 5, according to the corresponding focus type candidate collection of rule model and similarity mould
The corresponding focus type candidate collection of type, the corresponding focus type of the central issue extracted, comprising: step 502, seek rule
The intersection of the corresponding focus type candidate collection of model and the corresponding focus type candidate collection of similarity model；Step 504, work as friendship
When the number of focalization type is 1, using the focus type in intersection as the corresponding focus type of central issue extracted.When
When the number of focalization type being handed over to be greater than 1, the probability value and corresponding distance value of each focus type are obtained；It calculates each
The probability value of focus type and the product of respective distances value inverse, using the maximum focus type of product as the central issue extracted
Corresponding focus type.When handing over the number of focalization type to be 0, the probability value of each focus type is obtained；By probability value
Maximum focus type is as the corresponding focus type of central issue extracted.
Seek the corresponding focus type candidate collection of rule model and the corresponding focus type candidate collection of similarity model
Intersection, for example handing over focalization type is A, using focus type A as the corresponding focus type of central issue extracted.Work as intersection
When middle focus type number is greater than 1, for example handing over focalization type includes A and B, and the probability value of focus type A is 1/3, corresponding
Distance value be 1/2, the probability value of focus type B is 1/4, and corresponding distance value is 1/2.Calculate focus type probability value with it is right
The product for answering distance value inverse, obtaining the corresponding product of focus type A is 2/3, and the corresponding product of focus type B is 1/2, because
This, using focus type B as the corresponding focus type of central issue extracted.When handing over focalization type number is 0, such as
The corresponding focus type candidate collection of rule model includes A and B, and the probability value of focus type A is 1/3, the probability value of focus type B
1/4, the corresponding focus type candidate collection of similarity model includes C, using focus type A as the corresponding coke of central issue extracted
Vertex type.
In one embodiment, focal position mark is carried out to judgement document's sample according to focus locating rule, obtains band
Before judgement document's sample of mark, further includes: obtain the model generated by third party's open source library；By gibbs sampler with
And pseudo- likelihood algorithm, the model generated to third party's open source library optimize, the model after obtaining optimization processing；It is fixed according to focus
Position rule carries out focal position mark to judgement document's sample, obtains judgement document's sample with mark, comprising: at optimization
Model and focus locating rule after reason carry out focal position mark to judgement document's sample, obtain the judgement document with mark
Sample.Using the generation model in third party's open source library Snorkel, the generation model of Snorkel is the pre- of different label equations
It surveys result and regards the factor in factor graph as, reasonable value is therefrom extracted using gibbs sampler, and pass through pseudo- likelihood algorithm optimization
Entire graph model, the process do not need to provide true label value.
In one embodiment, according to disaggregated model and the list of types of central issue, to the central issue of extraction into
Row is classified, after the automatic classification results for the central issue extracted, further includes: is chosen from classified focus to artificial
Focus is verified, terminal will be sent to manual verification's focus and verified；Manual sort is received as a result, manual sort's result is by end
It treats manual verification's focus and is classified to obtain in end；Manual sort's result is compared with automatic classification results, is obtained automatic
The accuracy rate of classification results.Expert extracts the examination of 10-20 item from every class focus of automatic classification results, obtains manual sort's knot
Focus is divided into three classes by fruit according to manual sort's result: can reach the focus type of degree of precision using automatic classification, increase and divide
Classification can match correct focus type automatically after rule-like, and incompetent focus type of classifying automatically.Expert is newly-increased
Classifying rules, for example increase or modify keyword dictionary, some simple rules etc. that regular expression can be used to position, root are provided
It is optimized according to newly-increased classifying rules.It is right the step of duplication model optimization, category of model, expert's verifying for preceding two classes focus
It is then marked by professional mark personnel in latter class focus, until all focuses are all classified correctly.
It should be understood that although each step in the flow chart of Fig. 1-5 is successively shown according to the instruction of arrow,
These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these steps
Execution there is no stringent sequences to limit, these steps can execute in other order.Moreover, at least one in Fig. 1-5
Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-steps
Completion is executed, but can be executed at different times, the execution sequence in these sub-steps or stage is also not necessarily successively
It carries out, but can be at least part of the sub-step or stage of other steps or other steps in turn or alternately
It executes.
In one embodiment, as shown in fig. 6, providing a kind of mark sorter of central issue, comprising: rule obtains
Modulus block 602, sample labeling module 604, focus extraction module 606 and classification of focal spot module 608.Rule acquisition module is used for
Focus locating rule and judgement document's sample are obtained, focus locating rule is used to position the focal position of judgement document；Sample
Labeling module obtains the judge with mark for carrying out focal position mark to judgement document's sample according to focus locating rule
Document sample；Focus extraction module for extracting the central issue in judgement document's sample with mark, and obtains disaggregated model
And the list of types of central issue；Classification of focal spot module is right for the list of types according to disaggregated model and central issue
The central issue of extraction is classified, the automatic classification results for the central issue extracted.
In one embodiment, disaggregated model includes rule model and similarity model, and classification of focal spot module includes: rule
Then model focus type Candidate Set acquiring unit, for striving according to the list of types of central issue, rule model and extraction
Focus is discussed, the corresponding focus type candidate collection of rule model is obtained；Similarity model focus type Candidate Set acquiring unit, is used for
According to the list of types of central issue, similarity model and the central issue of extraction, the corresponding focus of similarity model is obtained
Type candidate collection；Focus type determining units, for according to the corresponding focus type candidate collection of rule model and similarity mould
The corresponding focus type candidate collection of type, the corresponding focus type of the central issue extracted.
In one embodiment, rule model focus type Candidate Set acquiring unit includes: focus processing unit, for leading to
It crosses rule model and redundancy word and the processing of shared word is removed to the central issue of extraction, and carry out synonym conversion process；Word
Allusion quotation generation unit counts central issue all types of in processed central issue for the list of types according to central issue,
Generate focus type keyword dictionary；Focus sentence processing unit, for obtaining the focus sentence in judgement document's sample, focus point sentence
It is removed redundancy word and synonym conversion process；Probability acquiring unit, for obtaining the focus in processed focus sentence
Focus descriptor is successively compared with focus type keyword dictionary, obtains each focus descriptor and belong to focus by descriptor
The probability value of central issue type in type keyword dictionary；Focus type generation unit, for traversing focus type keyword
Dictionary chooses probability value and is greater than focus type candidate collection of the central issue type of preset threshold as rule model.
In one embodiment, similarity model focus type Candidate Set acquiring unit includes: pretreatment unit, for leading to
It crosses similarity model and redundancy word and the processing of shared word is removed to the central issue extracted, and carry out at synonym conversion
Reason；Keyword generation unit counts all types of in processed central issue and strives for the list of types according to central issue
Focus is discussed, focus keyword is generated；Sentence vector generation unit, for each central issue in focus keyword to be converted to dispute
Focus term vector obtains focus type sentence vector using the mean value of central issue term vector as central issue sentence vector；Focus sentence
Focus sentence is converted to focus and describes sentence vector by converting unit for obtaining the focus sentence in judgement document's sample；Similarity meter
Calculate unit, for by focus describe sentence vector traversal focus type sentence vector, by similarity calculation obtain focus type away from
From value, selected distance value meets focus type candidate collection of the focus type as similarity model of preset condition.
In one embodiment, focus type determining units include: intersection acquiring unit, corresponding for seeking rule model
Focus type candidate collection and the corresponding focus type candidate collection of similarity model intersection；First judging unit, for working as
When the number of focalization type being handed over to be 1, using the focus type in intersection as the corresponding focus type of central issue extracted.
It in one embodiment, further include second judgment unit after intersection acquiring unit, for when friendship focalization class
When the number of type is greater than 1, the probability value and corresponding distance value of each focus type are obtained；Calculate the general of each focus type
The product of rate value and respective distances value inverse, using the maximum focus type of product as the corresponding focus class of central issue extracted
Type.
It in one embodiment, further include third judging unit after intersection acquiring unit, for when friendship focalization class
When the number of type is 0, the probability value of each focus type is obtained；The maximum focus type of probability value is burnt as the dispute extracted
The corresponding focus type of point.
In one embodiment, further include optimization module before sample labeling module, increased income for obtaining by third party
The model that library generates；By gibbs sampler and pseudo- likelihood algorithm, the model generated to third party's open source library is optimized, is obtained
Model after to optimization processing；Sample labeling module, for according to after optimization processing model and focus locating rule to sanction
Sentence document sample and carry out focal position mark, obtains judgement document's sample with mark.
It in one embodiment, further include authentication module after classification of focal spot module, for being selected from classified focus
It takes to manual verification's focus, terminal will be sent to manual verification's focus and verified；Manual sort is received as a result, manual sort
As a result manual verification's focus is treated by terminal to be classified to obtain；Manual sort's result is compared with automatic classification results,
Obtain the accuracy rate of automatic classification results.
The specific of mark sorter about central issue limits the mark that may refer to above for central issue
The restriction of classification method, details are not described herein.Modules in the mark sorter of above-mentioned central issue can whole or portion
Divide and is realized by software, hardware and combinations thereof.Above-mentioned each module can be embedded in the form of hardware or independently of computer equipment
In processor in, can also be stored in a software form in the memory in computer equipment, in order to processor calling hold
The corresponding operation of the above modules of row.
In one embodiment, a kind of computer equipment is provided, which can be terminal, internal structure
Figure can be as shown in Figure 7.The computer equipment includes processor, the memory, network interface, display connected by system bus
Screen and input unit.Wherein, the processor of the computer equipment is for providing calculating and control ability.The computer equipment is deposited
Reservoir includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system and computer journey
Sequence.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The network interface of machine equipment is used to communicate with external terminal by network connection.When the computer program is executed by processor with
Realize a kind of mark classification method of central issue.The display screen of the computer equipment can be liquid crystal display or electronic ink
Water display screen, the input unit of the computer equipment can be the touch layer covered on display screen, be also possible to computer equipment
Key, trace ball or the Trackpad being arranged on shell can also be external keyboard, Trackpad or mouse etc..
It will be understood by those skilled in the art that structure shown in Fig. 7, only part relevant to application scheme is tied
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment
It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored with
The step of computer program, which realizes the mark classification of central issue in any embodiment when executing computer program.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
The step of mark classification method of central issue in any embodiment is realized when machine program is executed by processor.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of mark classification method of central issue, which comprises
Focus locating rule and judgement document's sample are obtained, the focus locating rule is used to position the focus position of judgement document
It sets；
Focal position mark is carried out to judgement document's sample according to the focus locating rule, obtains judge's text with mark
Book sample；
The central issue in judgement document's sample of the band mark is extracted, and obtains the type of disaggregated model and central issue
List；
According to the disaggregated model and the list of types of the central issue, classifies to the central issue of extraction, obtain
The automatic classification results of the central issue of the extraction.
2. the method according to claim 1, wherein the disaggregated model includes rule model and similarity mould
Type, it is described according to the disaggregated model and the list of types of the central issue, classify to the central issue of extraction, obtains
To the automatic classification results of the central issue of the extraction, comprising:
According to the central issue of the list of types of the central issue, the rule model and extraction, the regular mould is obtained
The corresponding focus type candidate collection of type；
According to the central issue of the list of types of the central issue, the similarity model and extraction, obtain described similar
Spend the corresponding focus type candidate collection of model；
According to the corresponding focus type candidate collection of the rule model and the corresponding focus type candidate of the similarity model
Collection, the corresponding focus type of the central issue extracted.
3. according to the method described in claim 2, it is characterized in that, the list of types according to the central issue, described
Rule model and the central issue of extraction obtain the corresponding focus type candidate collection of the rule model, comprising:
It is removed redundancy word and the processing of shared word by central issue of the rule model to extraction, and carries out synonym and turns
Change processing；
According to the list of types of the central issue, central issue all types of in processed central issue is counted, is generated burnt
Vertex type keyword dictionary；
The focus sentence in judgement document's sample is obtained, the focus sentence is removed at redundancy word and synonym conversion
Reason；
Obtain the focus descriptor in processed focus sentence, by the focus descriptor successively with the focus type keyword
Dictionary is compared, and obtains the probability that each focus descriptor belongs to central issue type in the focus type keyword dictionary
Value；
The focus type keyword dictionary is traversed, probability value is chosen and is greater than the central issue type of preset threshold as regular mould
The focus type candidate collection of type.
4. according to the method described in claim 2, it is characterized in that, the list of types according to the central issue, described
Similarity model and the central issue of extraction obtain the corresponding focus type candidate collection of the similarity model, comprising:
Redundancy word is removed to the central issue extracted by similarity model and shared word is handled, and carries out synonym and turns
Change processing；
According to the list of types of the central issue, central issue all types of in processed central issue is counted, is generated burnt
Point keyword；
Each central issue in the focus keyword is converted into central issue term vector, with the mean value of central issue term vector
As central issue sentence vector, focus type sentence vector is obtained；
The focus sentence in judgement document's sample is obtained, the focus sentence is converted into focus and describes sentence vector；
The focus is described into sentence vector and traverses the focus type sentence vector, by similarity calculation obtain focus type away from
From value, selected distance value meets focus type candidate collection of the focus type as similarity model of preset condition.
5. according to the method described in claim 2, it is characterized in that, described according to the corresponding focus type marquis of the rule model
Selected works and the corresponding focus type candidate collection of the similarity model, the corresponding focus type of the central issue extracted,
Include:
Seek the corresponding focus type candidate collection of the rule model and the corresponding focus type candidate of the similarity model
The intersection of collection；
When the number for handing over focalization type is 1, using the focus type in the intersection as the central issue extracted
Corresponding focus type.
6. according to the method described in claim 5, it is characterized in that, described seek the corresponding focus type marquis of the rule model
After the intersection of selected works and the corresponding focus type candidate collection of the similarity model, further includes:
When it is described hand over focalization type number be greater than 1 when, obtain each focus type probability value and corresponding distance
Value；
The probability value of each focus type and the product of respective distances value inverse are calculated, the maximum focus type of product is made
For the corresponding focus type of central issue of extraction.
7. according to the method described in claim 5, it is characterized in that, described seek the corresponding focus type marquis of the rule model
After the intersection of selected works and the corresponding focus type candidate collection of the similarity model, further includes:
When the number for handing over focalization type is 0, the probability value of each focus type is obtained；
Using the maximum focus type of probability value as the corresponding focus type of central issue extracted.
8. a kind of mark sorter of central issue, which is characterized in that described device includes:
Rule acquisition module, for obtaining focus locating rule and judgement document's sample, the focus locating rule is for fixed
The focal position of position judgement document；
Sample labeling module, for carrying out focal position mark to judgement document's sample according to the focus locating rule,
Obtain judgement document's sample with mark；
Focus extraction module, the central issue in judgement document's sample for extracting the band mark, and obtain disaggregated model
And the list of types of central issue；
Classification of focal spot module, the dispute for the list of types according to the disaggregated model and the central issue, to extraction
Focus is classified, and the automatic classification results of the central issue of the extraction are obtained.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 7 the method when executing the computer program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 7 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910185041.5A CN109992664B (en) | 2019-03-12 | 2019-03-12 | Dispute focus label classification method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910185041.5A CN109992664B (en) | 2019-03-12 | 2019-03-12 | Dispute focus label classification method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109992664A true CN109992664A (en) | 2019-07-09 |
| CN109992664B CN109992664B (en) | 2023-04-18 |
Family
ID=67130621
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910185041.5A Active CN109992664B (en) | 2019-03-12 | 2019-03-12 | Dispute focus label classification method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109992664B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321439A (en) * | 2019-07-10 | 2019-10-11 | 北京市律典通科技有限公司 | A kind of electronics marking management method and system |
| CN110674633A (en) * | 2019-09-18 | 2020-01-10 | 平安科技（深圳）有限公司 | Document review proofreading method and device, storage medium and electronic equipment |
| CN110765266A (en) * | 2019-09-20 | 2020-02-07 | 成都星云律例科技有限责任公司 | Method and system for merging similar dispute focuses of referee documents |
| CN110825879A (en) * | 2019-09-18 | 2020-02-21 | 平安科技（深圳）有限公司 | Case decision result determination method, device and equipment and computer readable storage medium |
| CN110825872A (en) * | 2019-09-11 | 2020-02-21 | 成都数之联科技有限公司 | Method and system for extracting and classifying litigation request information |
| CN110889502A (en) * | 2019-10-15 | 2020-03-17 | 东南大学 | Deep learning-based dispute focus generation method |
| CN111159017A (en) * | 2019-12-17 | 2020-05-15 | 北京中科晶上超媒体信息技术有限公司 | Test case generation method based on slot filling |
| CN111814477A (en) * | 2020-07-06 | 2020-10-23 | 重庆邮电大学 | Dispute focus discovery method and device based on dispute focus entity and terminal |
| CN112487146A (en) * | 2020-12-02 | 2021-03-12 | 重庆邮电大学 | Legal case dispute focus acquisition method and device and computer equipment |
| CN112580338A (en) * | 2019-09-27 | 2021-03-30 | 北京国双科技有限公司 | Method and device for determining dispute focus, storage medium and equipment |
| CN113553856A (en) * | 2021-06-16 | 2021-10-26 | 吉林大学 | Deep neural network-based dispute focus identification method |
| CN113792545A (en) * | 2021-11-16 | 2021-12-14 | 成都索贝数码科技股份有限公司 | News event activity name extraction method based on deep learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766426A (en) * | 2017-09-14 | 2018-03-06 | 北京百分点信息科技有限公司 | A kind of file classification method, device and electronic equipment |
| CN108280149A (en) * | 2018-01-04 | 2018-07-13 | 东南大学 | A kind of doctor-patient dispute class case recommendation method based on various dimensions tag along sort |
| CN109033105A (en) * | 2017-06-09 | 2018-12-18 | 北京国双科技有限公司 | The method and apparatus for obtaining judgement document's focus |
| CN109033041A (en) * | 2017-06-09 | 2018-12-18 | 北京国双科技有限公司 | The treating method and apparatus of document similarity |
| 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 |
-
2019
- 2019-03-12 CN CN201910185041.5A patent/CN109992664B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109033105A (en) * | 2017-06-09 | 2018-12-18 | 北京国双科技有限公司 | The method and apparatus for obtaining judgement document's focus |
| CN109033041A (en) * | 2017-06-09 | 2018-12-18 | 北京国双科技有限公司 | The treating method and apparatus of document similarity |
| CN107766426A (en) * | 2017-09-14 | 2018-03-06 | 北京百分点信息科技有限公司 | A kind of file classification method, device and electronic equipment |
| CN108280149A (en) * | 2018-01-04 | 2018-07-13 | 东南大学 | A kind of doctor-patient dispute class case recommendation method based on various dimensions tag along sort |
| 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 |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110321439A (en) * | 2019-07-10 | 2019-10-11 | 北京市律典通科技有限公司 | A kind of electronics marking management method and system |
| CN110825872A (en) * | 2019-09-11 | 2020-02-21 | 成都数之联科技有限公司 | Method and system for extracting and classifying litigation request information |
| CN110825872B (en) * | 2019-09-11 | 2023-05-23 | 成都数之联科技股份有限公司 | Method and system for extracting and classifying litigation request information |
| CN110674633A (en) * | 2019-09-18 | 2020-01-10 | 平安科技（深圳）有限公司 | Document review proofreading method and device, storage medium and electronic equipment |
| CN110825879A (en) * | 2019-09-18 | 2020-02-21 | 平安科技（深圳）有限公司 | Case decision result determination method, device and equipment and computer readable storage medium |
| CN110825879B (en) * | 2019-09-18 | 2024-05-07 | 平安科技（深圳）有限公司 | Decide a case result determination method, device, equipment and computer readable storage medium |
| CN110765266A (en) * | 2019-09-20 | 2020-02-07 | 成都星云律例科技有限责任公司 | Method and system for merging similar dispute focuses of referee documents |
| CN110765266B (en) * | 2019-09-20 | 2022-07-22 | 成都星云律例科技有限责任公司 | Method and system for merging similar dispute focuses of referee documents |
| CN112580338A (en) * | 2019-09-27 | 2021-03-30 | 北京国双科技有限公司 | Method and device for determining dispute focus, storage medium and equipment |
| CN110889502B (en) * | 2019-10-15 | 2024-02-06 | 东南大学 | Deep learning-based dispute focus generation method |
| CN110889502A (en) * | 2019-10-15 | 2020-03-17 | 东南大学 | Deep learning-based dispute focus generation method |
| CN111159017A (en) * | 2019-12-17 | 2020-05-15 | 北京中科晶上超媒体信息技术有限公司 | Test case generation method based on slot filling |
| CN111814477A (en) * | 2020-07-06 | 2020-10-23 | 重庆邮电大学 | Dispute focus discovery method and device based on dispute focus entity and terminal |
| CN111814477B (en) * | 2020-07-06 | 2022-06-21 | 重庆邮电大学 | Dispute focus discovery method and device based on dispute focus entity and terminal |
| CN112487146A (en) * | 2020-12-02 | 2021-03-12 | 重庆邮电大学 | Legal case dispute focus acquisition method and device and computer equipment |
| CN112487146B (en) * | 2020-12-02 | 2022-05-31 | 重庆邮电大学 | Legal case dispute focus acquisition method and device and computer equipment |
| CN113553856A (en) * | 2021-06-16 | 2021-10-26 | 吉林大学 | Deep neural network-based dispute focus identification method |
| CN113792545A (en) * | 2021-11-16 | 2021-12-14 | 成都索贝数码科技股份有限公司 | News event activity name extraction method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109992664B (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109992664A (en) | Mark classification method, device, computer equipment and the storage medium of central issue | |
| CA3129721C (en) | Pre-trained contextual embedding models for named entity recognition and confidence prediction | |
| Zayats et al. | Disfluency detection using a bidirectional LSTM | |
| Dai et al. | Multimodal end-to-end sparse model for emotion recognition | |
| CN108376151A (en) | Question classification method, device, computer equipment and storage medium | |
| CN112231472B (en) | Judicial public opinion sensitive information identification method integrated with domain term dictionary | |
| CN108804414A (en) | Text modification method, device, smart machine and readable storage medium storing program for executing | |
| CN110489550A (en) | File classification method, device and computer equipment based on combination neural net | |
| CN111291570A (en) | Method and device for realizing element identification in judicial documents | |
| CN115310425B (en) | Policy text analysis method based on policy text classification and key information identification | |
| CN111046670B (en) | Entity and relationship combined extraction method based on drug case legal documents | |
| CN110442859B (en) | Labeling corpus generation method, device, equipment and storage medium | |
| CN109087205A (en) | Prediction technique and device, the computer equipment and readable storage medium storing program for executing of public opinion index | |
| CN111462752B (en) | Attention mechanism, feature embedding and BI-LSTM (business-to-business) based customer intention recognition method | |
| CN110750978A (en) | Emotional tendency analysis method and device, electronic equipment and storage medium | |
| CN106910512A (en) | The analysis method of voice document, apparatus and system | |
| CN113886562A (en) | AI resume screening method, system, equipment and storage medium | |
| CN115238697A (en) | Judicial named entity recognition method based on natural language processing | |
| CN116010902A (en) | Cross-modal fusion-based music emotion recognition method and system | |
| CN115185918A (en) | Method and device for automatically classifying system logs | |
| CN113868422A (en) | Multi-label inspection work order problem traceability identification method and device | |
| Hillebrand et al. | Towards automating numerical consistency checks in financial reports | |
| CN113486174A (en) | Model training, reading understanding method and device, electronic equipment and storage medium | |
| CN116910251A (en) | Text classification method, device, equipment and medium based on BERT model | |
| CN115994220A (en) | Contact net text data defect identification method and device based on semantic mining |
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 |