CN107330011A - The recognition methods of the name entity of many strategy fusions and device - Google Patents
The recognition methods of the name entity of many strategy fusions and device Download PDFInfo
- Publication number
- CN107330011A CN107330011A CN201710447439.2A CN201710447439A CN107330011A CN 107330011 A CN107330011 A CN 107330011A CN 201710447439 A CN201710447439 A CN 201710447439A CN 107330011 A CN107330011 A CN 107330011A
- Authority
- CN
- China
- Prior art keywords
- recognition result
- name entity
- recognition
- identification
- language material
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Character Discrimination (AREA)
Abstract
This application discloses a kind of recognition methods of name entity of many strategy fusions and device, the name entity in the language material obtained is recognized using the first identification model, obtain the first recognition result, in the method that the application is provided, first identification model can update and expand corpus, so as to identify the name entity newly produced in language material, and then first recognition result has higher accuracy rate, recycle the name entity in the method identification language material of many identification model fusions, obtain the second recognition result, merge first recognition result and the second recognition result obtains the 3rd recognition result, semantic digging system is recycled to carry out role's distribution to the 3rd recognition result, and export the name entity with role, it is achieved thereby that in data magnanimity, entity type variation, neologisms reliably identify name entity when emerging in an endless stream, and role's distribution is carried out to the name entity identified.
Description
Technical field
The application is related to natural language processing field, more particularly to a kind of recognition methods of the name entity of many strategy fusions
And device.
Background technology
Name entity is exactly name, mechanism name, place name and other all entities with entitled mark, during it is text
Basic information element, is the important carrier of information representation, is the basis of correct understanding and processing text message.Chinese name is real
Body identification is one of basic task in natural language processing field, and its main task is to identify that the name occurred in text is real
Body and significant numeral classifier phrase are simultaneously sorted out, mainly including name, place name, institution term, temporal expression, the date,
Numerical expression etc..
In terms of natural language processing research, name Entity recognition is in information retrieval, information extraction, machine translation and text
The application fields such as classification play an important role, and it can significantly increase information retrieval, abstract extraction, information extraction, machine translation
It is that the automatic knowledge that obtains is laid a good foundation from text with the performance of the application system such as text classification.Name Entity recognition accurate
The height of rate and recall rate, directly decides the performance of the language understanding overall process such as syntactic analysis, semantic analysis.
In recent ten years, domestic and foreign scholars have been inquired into and furtherd investigate extensively to the entity recognition techniques in text.But
With the rapid development of Internet, a large amount of random, multi-field text datas constantly increase, to the accurate of name Entity recognition
Rate and recall rate propose new requirement, in addition, market there is also a need for carrying out role's distribution to the name entity recognized, because
This, either caters to the market demand, or improves the accuracy rate and recall rate of identification, the recognition methods of name entity all need into
One step is improved.
Conventional name entity recognition method is divided into two major classes at present：One is rule-based and knowledge method, and two be base
In the method for statistics.Rule-based and knowledge method is a kind of method used earliest, and this method is simple, convenient, shortcoming
It is to need substantial amounts of manual observation, it is portable poor.Statistics-Based Method will be named Entity recognition to regard a classification as and be asked
Topic, using similar SVMs, the sorting technique such as Bayesian model；Name Entity recognition can also be regarded as a sequence simultaneously
Row mark problem, sequence labelling is obtained using machine learning such as HMM, maximum entropy Markov chain, condition random fields
Model.But the above method or exist to be difficult to meet random, the multi-field, texts that make rapid progress a large amount of at present are named
The problem of Entity recognition, or the accuracy rate and recall rate of identification are low.
Such as, Chinese patent CN201610943210.3 disclose a kind of name entity recognition method based on artificial intelligence and
Device, the function mould that this method is generated by using conditional random field models and according to the retrieval daily record in preset time period
Type, while being named Entity recognition to text to be identified.The defect of the program is default entity word in its second identification
The functional mode that converges is to obtain candidates all in text to be identified by methods such as dictionary, rule match first to name entity word
Converge, so judge its as name entity vocabulary confidence level height, due to rule method tend to rely on concrete syntax,
Field and text formatting, compilation process is time-consuming and easily produces mistake, and needs exper ienced linguist to complete,
And the coverage rate of dictionary is relatively low, therefore this method is difficult to meet to largely random, the multi-field, texts that make rapid progress enter at present
Row name Entity recognition.
For another example Chinese patent CN201510889318.4 discloses a kind of name Entity recognition side suitable for social networks
Method, this method obtains first instance probability distribution and the test of Training document in the First ray marking model using initial construction
After the second instance probability distribution of document, similarity feature is extracted from social network information, similarity feature is based on again afterwards
Training obtains the second sequence labelling model, and then is obtained carrying out sequence labelling to test document based on the second sequence labelling model
The recognition result of entity is named, the accuracy rate and recall rate of final this method are low, its F value recognized is only 33.19%.
Therefore, need that exploitation one kind copes with data scale magnanimity, entity type variation, neologisms emerge in an endless stream badly
New situation, with higher recall rate and accuracy rate, but also the name entity that can be obtained to identification carries out the life of role's distribution
Name entity recognition method and name entity recognition device.
The content of the invention
This application provides a kind of recognition methods of name entity of many strategy fusions and device, to solve to advise in data
In the case that mould magnanimity, entity type variation, neologisms emerge in an endless stream, accuracy rate and recall rate to naming Entity recognition
It is low, and can not be to naming the problem of entity carries out role's distribution.
In a first aspect, this application provides a kind of recognition methods of the name entity of many strategy fusions, the recognition methods
Including：
Obtain language material；
The name entity in the language material is recognized using the first identification model, the first recognition result is obtained；
The name entity in the language material is recognized using the second identification model, the second recognition result is obtained；
First recognition result and second recognition result are merged, the 3rd recognition result is obtained.
Alternatively, first identification model is conditional random field models.
Alternatively, the name entity in the identification language material using the first identification model, obtains the first recognition result
Before step, in addition to：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain first identification using CRF kits
Model.
Alternatively,
The name entity recognized using the second identification model in the language material, the step of obtaining the second recognition result is wrapped
Include：
The language material is recognized using at least two identification models, every kind of identification model respectively obtains a sub- recognition result,
Generate sub- recognition result list；
Judge whether the recognition result in the sub- recognition result list meets output condition, output second is known if meeting
Other result；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value is the mode of at least two identification model.
Alternatively, the name entity recognized using the first identification model in the language material, obtains the first recognition result
The step of include：
The language material is recognized using at least two identification models, every kind of identification model respectively obtains a sub- recognition result,
Generate sub- recognition result list；
Judge whether the recognition result in the sub- recognition result list meets output condition, output first is known if meeting
Other result；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value is the mode of at least two identification model.
Second identification model is conditional random field models；
Described using the second identification model identification language material, before the step of obtaining the second recognition result, in addition to：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain second identification using CRF kits
Model.
Fusion first recognition result is with second recognition result, and the step of obtaining three recognition results is wrapped
Include：
Judge whether first recognition result meets fusion conditions with second recognition result, merged if meeting,
And export the result after fusion, i.e. the 3rd recognition result；
Alternatively, the fusion conditions are that first recognition result has identical name with second recognition result
Entity.
Alternatively, also include after the 3rd recognition result is obtained：Using semantic digging system to the 3rd recognition result
Carry out role's distribution, name entity of the generation with role.
Alternatively, the role is assigned as using semantic digging system, to naming entity point in the 3rd recognition result
Not carry out role's mark, and respectively output with role name entity.
Alternatively, the semantic digging system includes regular expression and text.
Second aspect, the application also provides a kind of name entity recognition device of many strategy fusions, and the name entity is known
Other device includes,
Language material acquiring unit, for obtaining language material；
First recognition unit, for recognizing the name entity in the language material using the first identification model, obtains the first knowledge
Other result；
Second recognition unit, for recognizing the name entity in the language material using the second identification model, obtains the second knowledge
Other result；
Recognition result integrated unit, for merging first recognition result and second recognition result, obtains the 3rd
Recognition result.Alternatively, first identification model is conditional random field models.
Alternatively, first recognition unit also includes model training unit, and the model training unit is used for：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain first identification using CRF kits
Model.
Alternatively, second recognition unit includes following subelement：
Many strategy recognition units, it is every kind of for recognizing the name entity in the language material using at least two identification models
Identification model respectively obtains a sub- recognition result, generates sub- recognition result list；
Recognition result output unit, for judging whether the recognition result in the sub- recognition result list meets output bars
Part, the second recognition result is exported if meeting.
Alternatively, the output condition be in the sub- recognition result list, it is identical name entity number reach it is pre-
If value, wherein, the preset value is the mode of at least two identification model.
Alternatively, first recognition unit includes following subelement：
Many strategy recognition units, it is every kind of for recognizing the name entity in the language material using at least two identification models
Identification model respectively obtains a sub- recognition result, generates sub- recognition result list；
Recognition result output unit, for judging whether the recognition result in the sub- recognition result list meets output bars
Part, the first recognition result is exported if meeting；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value is the mode of at least two identification model.
Alternatively, second identification model is conditional random field models；
Also include model training unit in second recognition unit, the model training unit is used for：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain second identification using CRF kits
Model.
Alternatively, the recognition result integrated unit, for judging that first recognition result is tied with the described second identification
Whether fruit meets fusion conditions, is merged if meeting, and exports the result after fusion, i.e. the 3rd recognition result.
Alternatively, the fusion refers to increase on the basis of the first recognition result the name increased newly in the second recognition result
Entity；
Alternatively, the fusion conditions be the second recognition result in exist on the basis of the first recognition result increase newly name
Entity.
Alternatively, in addition to role's allocation unit, for being carried out using semantic digging system to the 3rd recognition result
Role distributes, name entity of the generation with role.
Alternatively, role's allocation unit is used for using semantic digging system, to being named in the 3rd recognition result
Entity carries out role's mark, and name entity of the output with role respectively respectively.
Alternatively, the semantic digging system includes regular expression and text.
Brief description of the drawings
In order to illustrate more clearly of the technical scheme of the application, letter will be made to the required accompanying drawing used in embodiment below
Singly introduce, it should be apparent that, for those of ordinary skills, without having to pay creative labor,
Other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 shows a kind of method flow of the name entity recognition method for many strategy fusions that the embodiment of the present application is provided
Figure；
Fig. 2 shows the method flow diagram for the conditional random field models that the embodiment of the present application is provided；
Fig. 3 shows the structural representation for the name entity recognition device that the embodiment of the present application is provided；
Fig. 4 shows the structural representation for the computer system 400 that the embodiment of the present application is provided；
Fig. 5 shows the accuracy rate, recall rate and F value result line charts of experimental example 1；
Fig. 6 shows the accuracy rate, recall rate and F value result line charts of experimental example 2.
Embodiment
It is described in detail, will with these explanations becomes more with advantage the characteristics of the application below by the application
To be clear, clear and definite.
Special word " exemplary " is meant " being used as example, embodiment or illustrative " herein.Here as " exemplary "
Illustrated any embodiment should not necessarily be construed as preferred or advantageous over other embodiments.Although each of embodiment is shown in the drawings
In terms of kind, but unless otherwise indicated, it is not necessary to accompanying drawing drawn to scale.
The application described below.
According to the first aspect of the application there is provided a kind of recognition methods of the name entity of many strategy fusions, first is utilized
Name entity in the language material that identification model identification is obtained, obtains the first recognition result, described in the method that the application is provided
First identification model can update and expand corpus, so as to identify the name entity newly produced in language material, Jin Ersuo
The first recognition result is stated with higher accuracy rate, the name in the method identification language material of many identification model fusions is recycled
Entity, obtains the second recognition result, merges first recognition result and the second recognition result obtains the 3rd recognition result, so that
Realize and reliably identify name entity when data magnanimity, entity type variation, neologisms emerge in an endless stream, appoint
Selection of land, recycles semantic digging system to carry out role's distribution to the 3rd recognition result, and exports the name entity with role, from
And role's distribution is carried out to the name entity identified.
Specifically, as shown in figure 1, the name entity recognition method includes：
S101 obtains language material；
S102 recognizes the name entity in the language material using the first identification model, obtains the first recognition result；
S103 recognizes the name entity in the language material using the second identification model, obtains the second recognition result；
S104 merges first recognition result and second recognition result, obtains the 3rd recognition result；
Optionally, in addition to S105 carries out role's distribution, generation to the 3rd recognition result using semantic digging system
Name entity with role.
In this application, the language material refers to the text for being used as training or identification.
In the application is a kind of preferred embodiment, first identification model is conditional random field models, i.e. CRF
Model (Conditional Random Fields, conditional random field models), it has counted global probability in statistics, done
Data are considered during normalization in global distribution, rather than only in local normalization, so as to avoid asking for marking bias
Topic.
In this application, as shown in Fig. 2 when the first identification model selects CRF models, knowing using the first identification model
Not described language material, obtains also including before the first recognition result：
S301 sets up corpus；
S302 carries out part-of-speech tagging and sequence labelling to the language material in the corpus；
Language material after mark as training data, is trained to obtain described first by S303 using CRF kits
Identification model.
In this application, the corpus refers to the set of the language material of identification name entity, e.g., the people for public security system
In name recognition method, language material stock is exactly notes set；For the language material stock in medical system name entity recognition method just
It is case set；Corpus without specific area can also use the set for the language material that reptile obtains from network.
In this application, the corpus of setting up is imported including language material, the language material imported in above-mentioned corpus.
In this application, the language material in the corpus is processed into the form that can be recognized by CRF first, i.e. to language
Material progress part-of-speech tagging and sequence labelling, the training text string and test text string obtained, wherein, the training text after mark
This string is as training data, and the test text string after mark is used as test data.
In this application, when to CRF model trainings, the special characteristic of training data, then root are obtained according to feature templates
It is trained according to special characteristic, part-of-speech tagging and sequence labelling result, obtains CRF models, above and below the special characteristic includes
Literary feature, part of speech feature etc..
In this application, training result is tested using test data after being finished to CRF model trainings, when identification knot
When the F values of fruit are below 0.8, training data and test data are reacquired, continues to train, training uses what is newly obtained after finishing
Test data is tested, and when the F values of recognition result are less than 0.8, is repeated the above steps, until the F values of training result reach
More than 0.8, deconditioning, so as to obtain the first identification model.
In the present embodiment, the name entity indicia that the first identification model identification is obtained has first position information.
In the present embodiment, the name entity recognized using the second identification model in the language material, obtains second
The step of recognition result, includes：
The language material is recognized using at least two identification models, every kind of identification model respectively obtains a sub- recognition result,
Generate sub- recognition result list；
Judge whether the recognition result in the sub- recognition result list meets output condition, output second is known if meeting
Other result；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value is the mode of at least two identification model.
In this application, at least two identification model includes participle model and Named Entity Extraction Model.
In this application, the participle model includes nGram participle models (single order Markov Chain), HMM participle models
(HMM), the participle model with new word discovery function.
In this application, the name physical model includes the Named Entity Extraction Model based on maximum entropy, based on structure
Change the Named Entity Extraction Model of perceptron.
In this application, the nGram participle models obtain nGram statistical information by statistics first, then basis
The statistical information carries out participle to the language material for needing to recognize name entity, and this method can be looked after and is possible to, but can also be made
Index entry increase, such as " coming into search engine " can be divided into by 2-gram participles：Come into, enter to search, search for, indexing, engine.
In this application, the HMM participle models obtain HMM parameters, so by the participle training set marked
The language material for needing to recognize name entity is explained using viterbi algorithm afterwards, word segmentation result is obtained, it is independent that the model is based on output
Property is not it is assumed that consider contextual feature.
In this application, the participle model with new word discovery function is found by the identification model of rule or statistics
Name entity in language material, but relatively depend on training corpus.
In this application, the Named Entity Extraction Model based on maximum entropy results in all constraintss that meet
The model of Information Entropy Maximal in model, and can be by setting constraints to adjust model to the fitness of unknown data and right
The fitting degree of given data, again, the problem of it can also solve parameter smoothing in statistical model naturally.But the model
The timely air switch pin of calculation cost is larger, and Sparse Problem is than more serious.
In this application, feature extraction considers the overall situation in the Named Entity Extraction Model based on structuring perceptron
Structuring output so that model can carry out the overall situation structuring study.
In the present embodiment, the name entity that at least two identification model is identified is marked with the second place respectively
Information.
In this application, the output condition is that in the sub- recognition result list, the number of identical name entity reaches
To preset value, wherein, it is whether identical with the whether identical name entity for judging that various identification models are identified of second place information,
The preset value is the mode of at least two identification model.
Therefore, the recognition result obtained by above-mentioned model is merged, the intrinsic deficiency of each model itself can be made up so that is known
Other result is optimal.
In this application, described is to be determined by the F values of experimental result, as shown in the application experimental example 1, when using essence
Quasi- segmentation methods (combining language model, sequence labelling and HMM), the participle with new word discovery function are calculated
During the name entity identification algorithms of method and structuring perceptron, mode takes 3, as a result optimal.
Applicants have discovered that, judge whether recognition result, energy in the output sub- recognition result list using output condition
The enough farthest other result of deletion misrecognition, such as wrong identification, so as to improve the recall rate of final recognition result.
Applicants have discovered that, the language material is recognized using at least two identification models, name can be more accurately identified
Entity, so that multiple weak identification models are combined into one strong identification model, is supplemented basic result, and then improves identification
As a result.
In the application another preferred embodiment, the life recognized using the first identification model in the language material
Name entity, the step of obtaining the first recognition result can also be：
The language material is recognized using at least two identification models, every kind of identification model respectively obtains a sub- recognition result,
Generate sub- recognition result list；
Judge whether the recognition result in the sub- recognition result list meets output condition, output first is known if meeting
Other result；
In the present embodiment, the name entity that at least two identification model is identified is marked with first position respectively
Information.
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value whether identical with the whether identical name entity for judging that various identification models are identified of first position information
For the mode of at least two identification model.
In the present embodiment, second identification model is conditional random field models, preferably conditional random field models.
In the present embodiment, mark has information on second recognition result.
In this application, fusion first recognition result and second recognition result, obtain the 3rd identification knot
The step of fruit, includes：
Judge whether first recognition result meets fusion conditions with second recognition result, merged if meeting,
And export the result after fusion, i.e. the 3rd recognition result.
Applicants have discovered that, the first recognition result is merged with the second recognition result Ji Wei removing the first recognition result and the
The name entity repeated in two recognition results, so as to avoid the redundancy of data, and then improves the accuracy rate of identification and recalls
Rate.
In this application, the fusion refers to increase on the basis of the first recognition result what is increased newly in the second recognition result
Name entity.
In this application, the fusion conditions be the second recognition result in exist on the basis of the first recognition result increase newly
Name entity.
In the application is a kind of preferred embodiment, judge whether second place information and first position information are identical,
If it is different, the name entity increased newly in then judging the name entity for the second recognition result.
Alternatively, the semantic digging system, to naming entity to carry out role's mark respectively in the 3rd recognition result,
And output has the name entity of role respectively.
In this application, the semantic digging system can not only carry out role's distribution, additionally it is possible to name Entity recognition
As a result judged, determine whether it is name entity.
The semantic digging system includes regular expression and text.
For the recognition methods for the name entity for being more fully understood by many strategy fusions described herein, one is set forth below
Specific embodiment is illustrated.
Set up corpus.
Part-of-speech tagging and sequence labelling, wherein sequence mark are carried out to each subordinate sentence in the language material in corpus, i.e. language material
The corresponding word of entity will be named to be labeled with B, M, E during note, remaining word is marked with S, the training text string of acquisition.Assuming that one
Training text string is " check to have in discovery satchel through people's police and three see identity card perhaps ", and annotation results are as shown in table 1.
The text string of table 1 marks example
Using the corresponding annotation results of a large amount of training text strings as training data, it is trained using CRF.
Assuming that the user's input language material being currently received is " victim Ni Chengang alarms claim to find mobile phone not in Qinghe Oak Tree gulf
See ".The CRF models obtained using preceding step are named Entity recognition to user input language material, can be named
Entity " Ni Chengang ".
The method learnt afterwards using a variety of method integrations carries out supplement amendment to CRF results, and such as accurate word segmentation result will
Name Entity recognition in upper example is " Ni Chen ", and structuring perceptron recognition result is " Ni Chengang ", with new word discovery function
Recognition result is " Ni Chengang ", and mode is taken to the recognition result of several method, it may be determined that name Entity recognition result is " Ni Chen
Just ", rather than " Ni Chen ".
By regular expression present in semantic digging system or text, such as " victim's alarm ", on the one hand it can determine
" Ni Chengang " is correct name Entity recognition result, on the other hand can by role be defined as " victim ".
According to the second aspect of the application, as shown in figure 3, additionally providing a kind of name Entity recognition dress of many strategy fusions
Put, the name entity recognition device of many strategy fusions includes,
Language material acquiring unit 201, for obtaining language material；
First recognition unit 202, for recognizing the name entity in the language material using the first identification model, obtains first
Recognition result；
Second recognition unit 203, for recognizing the name entity in the language material using the second identification model, obtains second
Recognition result；
Recognition result integrated unit 204, for merging first recognition result and second recognition result, obtains
Three recognition results；
Optionally, in addition to role's allocation unit 205, for being entered using semantic digging system to the 3rd recognition result
Row role distributes, name entity of the generation with role.
In a kind of optional embodiment of the application, first identification model is conditional random field models.
Alternatively, first recognition unit also includes model training unit, and the model training unit is used for：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain first identification using CRF kits
Model.
Alternatively, second recognition unit includes following subelement：
Many strategy recognition units, it is every kind of for recognizing the name entity in the language material using at least two identification models
Identification model respectively obtains a sub- recognition result, generates sub- recognition result list；
Recognition result output unit, for judging whether the recognition result in the sub- recognition result list meets output bars
Part, the second recognition result is exported if meeting；
Alternatively, the output condition be in the sub- recognition result list, it is identical name entity number reach it is pre-
If value, wherein, the preset value is the mode of at least two identification model.
In another optional embodiment of the application, first recognition unit includes following subelement：
Many strategy recognition units, it is every kind of for recognizing the name entity in the language material using at least two identification models
Identification model respectively obtains a sub- recognition result, generates sub- recognition result list；
Recognition result output unit, for judging whether the recognition result in the sub- recognition result list meets output bars
Part, the first recognition result is exported if meeting；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, its
In, the preset value is the mode of at least two identification model.
Alternatively, second identification model is conditional random field models；
Also include model training unit in second recognition unit, the model training unit is used for：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain second identification using CRF kits
Model.
Alternatively, the recognition result integrated unit, for judging that first recognition result is tied with the described second identification
Whether fruit meets fusion conditions, is merged if meeting, and exports the result after fusion, i.e. the 3rd recognition result.
Alternatively, the fusion conditions are that the second recognition result and the first recognition result have identical name entity.
Alternatively, role's allocation unit is used for using semantic digging system, to being named in the 3rd recognition result
Entity carries out role's mark, and name entity of the output with role respectively respectively.
Alternatively, the semantic digging system includes regular expression and text.
Fig. 4 show can thereon implement embodiment computer system 400 block diagram.Computer system 400 is wrapped
Include processor 410, storage medium 420, system storage 430, monitor 440, keyboard 450, mouse 460, the and of network interface 420
Video adapter 480.These parts are coupled by system bus 490.
Storage medium 420 (such as hard disk) stores multiple programs, including operating system, application program and other program moulds
Block.User can input into computer system 400 order and information by input equipment, input equipment be, for example, keyboard 450,
Touch pad (not shown) and mouse 460.Text and graphical information are shown using monitor 440.
Operating system is on processor 410 and for coordinating and providing in the personal computer system 400 in Fig. 6
Various parts control.Furthermore, it is possible in computer system 400 using computer program with implement it is above-mentioned it is various implement
Example.
It would be recognized that hardware component shown in Fig. 4 is only for illustrative purposes, and physical unit may be according to be real
The computing device applying the application and dispose and change.
In addition, computer system 400 for example can be desktop computer, server computer, laptop computer or nothing
Line equipment, such as mobile phone, personal digital assistant (PDA), handheld computer.
The embodiment provides a kind of effective ways that name entity is extracted in the case of given document collected works.Implement
Example solve from the webpage typically organized with least cost extract any types entity the problem of.The weighting name entity proposed
Figure can be encoded to the complex relationship between each name entity and the type of other entities, therefore propagate seed on the diagram
Confidence level can make up the shortage of network size redundancy, and effective size of the organization can be supported to extract.Furthermore, it is possible to will life
Confidence spread on name sterogram is transformed into efficient matrix computations, and it can support the high efficiency extraction on extensive collected works.
It would be recognized that the embodiment in the range of the application can be embodied as to the form of computer program product, computer
Program product includes computer executable instructions, such as program code, and it can run on any with reference to appropriate operating system
In appropriate computing environment, operating system is, for example, Microsoft Windows, Linux or UNIX operating system.The application scope
Interior embodiment can also include program product, and program product includes computer-readable medium can for carrying or storing computer
Execute instruction or data structure are thereon.Such computer-readable medium can be it is any can by it is universal or special calculate
The usable medium that machine is accessed.For example, such computer-readable medium can include RAM, ROM, EPROM, EEPROM, CD-
ROM, magnetic disk storage or other storage devices, or can be used in carrying with form of computer-executable instructions or store desired
Program code and any other medium that can be accessed by universal or special computer.
Experimental example
Influence of the mode value to F values when experimental example 1 second is recognized
Used in the second identification step during the second identification in this experimental example, preset value is different, final name entity
Recognition result significant difference, this experimental example has investigated influence of the preset value to name Entity recognition result.
The preset value is the mode of at least two identification model；
The name Entity recognition result is weighed by F values, and F values are higher, and recognition result is more reliable, wherein,
The name entity number of accuracy rate (P)=correct number/machine recognition of name Entity recognition,
Name entity number in recall rate (R)=correct number/testing material of name Entity recognition.
F values=2*P*R/ (P+R).
Identification model used during the second identification includes accurate segmentation methods, with new word discovery function in this experimental example
The name entity identification algorithms of segmentation methods, structuring perceptron, wherein,
Accurate participle is the segmentation methods of a kind of combination language model, sequence labelling and HMM, it is preferable that
Thick cutting is carried out first by N-gram and HMM, CRF is then reused and fritter point；
Segmentation methods with new word discovery function find the neologisms in text by the identification model of rule or statistics；
The problem of structuring perceptron is used to solve sequence labelling.
The result of this experimental example as shown in Fig. 5 and table 1,
Influence of the preset value of table 1 to name Entity recognition result
In Figure 5, broken line A is the corresponding recall rate broken line of each preset value；Broken line B shows the corresponding F values folding of each preset value
Line；Broken line C is the corresponding accuracy rate broken line of each preset value.
From Fig. 5 and table 1, in this experimental example, when mode value is 3, F values reach maximum.
Entity recognition result is named when each identification model of experimental example 2 is used alone
A kind of result of identification model to name Entity recognition is used alone in the test of this experimental example, to contrast single identification mould
Type merges the reliability of two kinds of name entity recognition methods with many identification models.
Identification model used is respectively CRF identification models used in preliminary identification, the second identification in this experimental example
The middle accurate segmentation methods used, the segmentation methods with new word discovery function, the name Entity recognition of structuring perceptron are calculated
Method, as a result as shown in Fig. 6 and table 2.
The single identification model of table 2 names the reliability of entity recognition method
In figure 6, broken line A is the corresponding recall rate broken line of each recognition methods；Broken line B shows the corresponding F of each recognition methods
It is worth broken line；Broken line C is the corresponding accuracy rate broken line of each recognition methods.
From Fig. 6 and table 2, the name entity recognition method merged with many identification models (name of i.e. many strategy fusions
Entity recognition method) and (experimental example 1, mode be 3 result) compare, single identification model name entity recognition method F values compared with
It is low, i.e. the name Entity recognition result that the name entity recognition method of many identification models fusion provided with the application is obtained is more
To be reliable and stably.
The name Entity recognition result of each identification model of the application method of experimental example 3
This experimental example utilizes the method that the application is provided, and the first recognition result, the second recognition result and the 3rd are calculated respectively
Accuracy rate, recall rate and the F values of recognition result, it is as a result as shown in table 3 below.
The name Entity recognition result of each identification model of the application method of table 3
As shown in Table 3, the method provided according to the application, on the basis of the first recognition result and the second recognition result
The 3rd recognition result arrived, its accuracy rate, recall rate and F values have raising by a relatively large margin, i.e. the method that the application is provided
The new situations such as data scale magnanimity, entity type are diversified, neologisms emerge in an endless stream are coped with, with higher recall rate and standard
True rate.
The name entity recognition method and identifying device of many strategy fusions provided according to the application, with following beneficial effect
Really：
(1) scheme that the application is provided can be named entity knowledge by preliminary identification step to new data or frontier
Not, so as to adapt to when data scale magnanimity, entity type variation, neologisms emerge in an endless stream to name Entity recognition
Demand；
(2) second identification steps name the fusion of entity recognition method by many identification models, by multiple weak identification models
One strong identification model is combined into, the first recognition result is supplemented, so as to improve recognition result accuracy rate and recall rate；
(3) the name entity obtained using semantic digging system to the second identification carries out role's mark, so as to obtain role
Name entity after distribution；
(4) method that the application is provided, which can be migrated easily into new data and frontier, uses；
(5) method that the application is provided has higher accuracy rate and recall rate, and its F value is up to more than 0.8.
The application is described in detail above in association with embodiment and exemplary example, but these explanations are simultaneously
It is not intended that the limitation to the application.It will be appreciated by those skilled in the art that in the case of without departing from the application spirit and scope,
A variety of equivalencings, modification can be carried out to technical scheme and embodiments thereof or is improved, these each fall within the application
In the range of.The protection domain of the application is determined by the appended claims.
Claims (10)
1. a kind of recognition methods of the name entity of many strategy fusions, it is characterised in that including：
Obtain language material；
The name entity in the language material is recognized using the first identification model, the first recognition result is obtained；
The name entity in the language material is recognized using the second identification model, the second recognition result is obtained；
First recognition result and second recognition result are merged, the 3rd recognition result is obtained.
2. recognition methods according to claim 1, it is characterised in that
First identification model is conditional random field models；
Before name entity in the identification language material using the first identification model, the step of obtaining the first recognition result, also
Including：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained using CRF kits, obtains first identification model.
3. recognition methods according to claim 2, it is characterised in that described to recognize the language material using the second identification model
In name entity, the step of obtaining the second recognition result include：
The name entity in the language material is recognized using at least two identification models, every kind of identification model respectively obtains a son knowledge
Other result, generates sub- recognition result list；
Judge whether the recognition result in the sub- recognition result list meets output condition, the identification of output second knot if meeting
Really；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, wherein, institute
State the mode that preset value is at least two identification model.
4. recognition methods according to claim 3, it is characterised in that fusion first recognition result and described the
Two recognition results, the step of obtaining three recognition results includes：
Judge whether first recognition result meets fusion conditions with second recognition result, merged if meeting, and it is defeated
The result gone out after fusion, i.e. the 3rd recognition result；
The fusion refers to increase on the basis of the first recognition result the name entity increased newly in the second recognition result；
The fusion conditions be the second recognition result in exist on the basis of the first recognition result increase newly name entity.
5. recognition methods according to claim 1, it is characterised in that also include after the 3rd recognition result is obtained：
Role's distribution is carried out to the 3rd recognition result using semantic digging system, generation has the name entity of role, its
In,
The role is assigned as using semantic digging system, to naming entity to carry out role's mark respectively in the 3rd recognition result
Note, and name entity of the output with role respectively；
The semantic digging system includes regular expression and text.
6. a kind of name entity recognition device of many strategy fusions, it is characterised in that the name entity recognition device includes,
Language material acquiring unit, for obtaining language material；
First recognition unit, for recognizing the name entity in the language material using the first identification model, obtains the first identification knot
Really；
Second recognition unit, for recognizing the name entity in the language material using the second identification model, obtains the second identification knot
Really；
Recognition result integrated unit, for merging first recognition result and second recognition result, obtains the 3rd identification
As a result.
7. identifying device according to claim 6, it is characterised in that
First identification model is conditional random field models；
Also include model training unit in first recognition unit, the model training unit is used for：
Set up corpus；
Part-of-speech tagging and sequence labelling are carried out to the language material in the corpus；
Using the language material after mark as training data, it is trained to obtain first identification model using CRF kits.
8. identifying device according to claim 7, it is characterised in that second recognition unit includes following subelement：
Many strategy recognition units, for recognizing the name entity in the language material, every kind of identification using at least two identification models
Model respectively obtains a sub- recognition result, generates sub- recognition result list；
Recognition result output unit, for judging whether the recognition result in the sub- recognition result list meets output condition,
The second recognition result is exported if meeting；
The output condition is that in the sub- recognition result list, the number of identical name entity reaches preset value, wherein, institute
State the mode that preset value is at least two identification model.
9. identifying device according to claim 8, it is characterised in that
The recognition result integrated unit, melts for judging first recognition result with whether second recognition result meets
Conjunction condition, is merged if meeting, and exports the result after fusion, i.e. the 3rd recognition result；
The fusion refers to increase on the basis of the first recognition result the name entity increased newly in the second recognition result；
The fusion conditions be the second recognition result in exist on the basis of the first recognition result increase newly name entity.
10. identifying device according to claim 6, it is characterised in that also including role's allocation unit, for utilizing semanteme
Digging system carries out role's distribution to the 3rd recognition result, and generation has the name entity of role, wherein,
Role's allocation unit is used for using semantic digging system, to naming entity to carry out respectively in the 3rd recognition result
Role marks, and name entity of the output with role respectively；
The semantic digging system includes regular expression and text.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710447439.2A CN107330011B (en) | 2017-06-14 | 2017-06-14 | The recognition methods of the name entity of more strategy fusions and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710447439.2A CN107330011B (en) | 2017-06-14 | 2017-06-14 | The recognition methods of the name entity of more strategy fusions and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107330011A true CN107330011A (en) | 2017-11-07 |
| CN107330011B CN107330011B (en) | 2019-03-26 |
Family
ID=60195026
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710447439.2A Active CN107330011B (en) | 2017-06-14 | 2017-06-14 | The recognition methods of the name entity of more strategy fusions and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107330011B (en) |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108108350A (en) * | 2017-11-29 | 2018-06-01 | 北京小米移动软件有限公司 | Name word recognition method and device |
| CN108170674A (en) * | 2017-12-27 | 2018-06-15 | 东软集团股份有限公司 | Part-of-speech tagging method and apparatus, program product and storage medium |
| CN108363701A (en) * | 2018-04-13 | 2018-08-03 | 达而观信息科技（上海）有限公司 | Name entity recognition method and system |
| CN108388638A (en) * | 2018-02-26 | 2018-08-10 | 出门问问信息科技有限公司 | Semantic analytic method, device, equipment and storage medium |
| CN108829681A (en) * | 2018-06-28 | 2018-11-16 | 北京神州泰岳软件股份有限公司 | A kind of name entity extraction method and device |
| CN109086274A (en) * | 2018-08-23 | 2018-12-25 | 电子科技大学 | English social media short text time expression recognition method based on restricted model |
| CN109543153A (en) * | 2018-11-13 | 2019-03-29 | 成都数联铭品科技有限公司 | A kind of sequence labelling system and method |
| CN109791570A (en) * | 2018-12-13 | 2019-05-21 | 香港应用科技研究院有限公司 | Efficiently and accurately name entity recognition method and device |
| CN109815296A (en) * | 2018-12-29 | 2019-05-28 | 北京中科闻歌科技股份有限公司 | The personage's construction of knowledge base method, apparatus and storage medium of notarization document |
| CN109886270A (en) * | 2019-01-17 | 2019-06-14 | 大连理工大学 | A kind of case element recognition methods towards electronics folder notes text |
| CN110489727A (en) * | 2019-07-12 | 2019-11-22 | 深圳追一科技有限公司 | Name recognition methods and relevant apparatus |
| CN110569332A (en) * | 2019-09-09 | 2019-12-13 | 腾讯科技（深圳）有限公司 | Sentence feature extraction processing method and device |
| CN110688467A (en) * | 2019-08-23 | 2020-01-14 | 北京百度网讯科技有限公司 | Named entity recognition method and device, computer equipment and storage medium |
| CN110750990A (en) * | 2019-09-17 | 2020-02-04 | 平安科技（深圳）有限公司 | Entity identification corpus labeling method, system, device and storage medium |
| CN110750991A (en) * | 2019-09-18 | 2020-02-04 | 平安科技（深圳）有限公司 | Entity identification method, device, equipment and computer readable storage medium |
| CN111125438A (en) * | 2019-12-25 | 2020-05-08 | 北京百度网讯科技有限公司 | Entity information extraction method and device, electronic equipment and storage medium |
| CN111178073A (en) * | 2018-10-23 | 2020-05-19 | 北京嘀嘀无限科技发展有限公司 | Text processing method and device, electronic equipment and storage medium |
| CN111178075A (en) * | 2019-12-19 | 2020-05-19 | 厦门快商通科技股份有限公司 | Online customer service log analysis method, device and equipment |
| CN111368541A (en) * | 2018-12-06 | 2020-07-03 | 北京搜狗科技发展有限公司 | Named entity identification method and device |
| CN111382570A (en) * | 2018-12-28 | 2020-07-07 | 深圳市优必选科技有限公司 | Text entity recognition method and device, computer equipment and storage medium |
| CN111400429A (en) * | 2020-03-09 | 2020-07-10 | 北京奇艺世纪科技有限公司 | Text entry searching method, device, system and storage medium |
| CN111488737A (en) * | 2019-01-09 | 2020-08-04 | 阿里巴巴集团控股有限公司 | Text recognition method, device and equipment |
| CN111797629A (en) * | 2020-06-23 | 2020-10-20 | 平安医疗健康管理股份有限公司 | Medical text data processing method and device, computer equipment and storage medium |
| WO2020215456A1 (en) * | 2019-04-26 | 2020-10-29 | 网宿科技股份有限公司 | Text labeling method and device based on teacher forcing |
| CN112270173A (en) * | 2020-10-27 | 2021-01-26 | 北京百度网讯科技有限公司 | Character mining method and device in text, electronic equipment and storage medium |
| EP3748548A4 (en) * | 2019-04-26 | 2021-03-10 | Wangsu Science & Technology Co., Ltd. | Adversarial learning-based text annotation method and device |
| CN112541065A (en) * | 2020-12-11 | 2021-03-23 | 浙江汉德瑞智能科技有限公司 | Medical new word discovery processing method based on representation learning |
| CN113051918A (en) * | 2019-12-26 | 2021-06-29 | 北京中科闻歌科技股份有限公司 | Named entity identification method, device, equipment and medium based on ensemble learning |
| CN113127060A (en) * | 2021-04-09 | 2021-07-16 | 中通服软件科技有限公司 | Software function point identification method based on natural language pre-training model (BERT) |
| CN113127645A (en) * | 2021-04-09 | 2021-07-16 | 厦门渊亭信息科技有限公司 | Automatic extraction method of large-scale knowledge graph ontology, terminal equipment and storage medium |
| CN113971216A (en) * | 2021-10-22 | 2022-01-25 | 北京百度网讯科技有限公司 | Data processing method and device, electronic equipment and memory |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110035210A1 (en) * | 2009-08-10 | 2011-02-10 | Benjamin Rosenfeld | Conditional random fields (crf)-based relation extraction system |
| CN102033879A (en) * | 2009-09-27 | 2011-04-27 | 腾讯科技（深圳）有限公司 | Method and device for identifying Chinese name |
| CN103309926A (en) * | 2013-03-12 | 2013-09-18 | 中国科学院声学研究所 | Chinese and English-named entity identification method and system based on conditional random field (CRF) |
| CN104572631A (en) * | 2014-12-03 | 2015-04-29 | 北京捷通华声语音技术有限公司 | Training method and system for language model |
| CN104933152A (en) * | 2015-06-24 | 2015-09-23 | 北京京东尚科信息技术有限公司 | Named entity recognition method and device |
| CN106202255A (en) * | 2016-06-30 | 2016-12-07 | 昆明理工大学 | Merge the Vietnamese name entity recognition method of physical characteristics |
| CN106326206A (en) * | 2015-06-24 | 2017-01-11 | 北京京东尚科信息技术有限公司 | Entity extraction method based on grammar templates |
| CN106503192A (en) * | 2016-10-31 | 2017-03-15 | 北京百度网讯科技有限公司 | Name entity recognition method and device based on artificial intelligence |
| CN106570132A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Document vector learning method with fusion of mentioned entity information |
| CN106649272A (en) * | 2016-12-23 | 2017-05-10 | 东北大学 | Named entity recognizing method based on mixed model |
-
2017
- 2017-06-14 CN CN201710447439.2A patent/CN107330011B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110035210A1 (en) * | 2009-08-10 | 2011-02-10 | Benjamin Rosenfeld | Conditional random fields (crf)-based relation extraction system |
| CN102033879A (en) * | 2009-09-27 | 2011-04-27 | 腾讯科技（深圳）有限公司 | Method and device for identifying Chinese name |
| CN103309926A (en) * | 2013-03-12 | 2013-09-18 | 中国科学院声学研究所 | Chinese and English-named entity identification method and system based on conditional random field (CRF) |
| CN104572631A (en) * | 2014-12-03 | 2015-04-29 | 北京捷通华声语音技术有限公司 | Training method and system for language model |
| CN104933152A (en) * | 2015-06-24 | 2015-09-23 | 北京京东尚科信息技术有限公司 | Named entity recognition method and device |
| CN106326206A (en) * | 2015-06-24 | 2017-01-11 | 北京京东尚科信息技术有限公司 | Entity extraction method based on grammar templates |
| CN106202255A (en) * | 2016-06-30 | 2016-12-07 | 昆明理工大学 | Merge the Vietnamese name entity recognition method of physical characteristics |
| CN106570132A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Document vector learning method with fusion of mentioned entity information |
| CN106503192A (en) * | 2016-10-31 | 2017-03-15 | 北京百度网讯科技有限公司 | Name entity recognition method and device based on artificial intelligence |
| CN106649272A (en) * | 2016-12-23 | 2017-05-10 | 东北大学 | Named entity recognizing method based on mixed model |
Non-Patent Citations (1)
| Title |
|---|
| 栾悉道: "《多媒体情报处理技术》", 31 May 2016, 国防工业出版社 * |
Cited By (50)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108108350A (en) * | 2017-11-29 | 2018-06-01 | 北京小米移动软件有限公司 | Name word recognition method and device |
| CN108170674A (en) * | 2017-12-27 | 2018-06-15 | 东软集团股份有限公司 | Part-of-speech tagging method and apparatus, program product and storage medium |
| CN108388638A (en) * | 2018-02-26 | 2018-08-10 | 出门问问信息科技有限公司 | Semantic analytic method, device, equipment and storage medium |
| CN108388638B (en) * | 2018-02-26 | 2020-09-18 | 出门问问信息科技有限公司 | Semantic parsing method, device, equipment and storage medium |
| CN108363701A (en) * | 2018-04-13 | 2018-08-03 | 达而观信息科技（上海）有限公司 | Name entity recognition method and system |
| CN108829681A (en) * | 2018-06-28 | 2018-11-16 | 北京神州泰岳软件股份有限公司 | A kind of name entity extraction method and device |
| CN108829681B (en) * | 2018-06-28 | 2022-11-11 | 鼎富智能科技有限公司 | Named entity extraction method and device |
| CN109086274A (en) * | 2018-08-23 | 2018-12-25 | 电子科技大学 | English social media short text time expression recognition method based on restricted model |
| CN109086274B (en) * | 2018-08-23 | 2020-06-26 | 电子科技大学 | English social media short text time expression recognition method based on constraint model |
| CN111178073A (en) * | 2018-10-23 | 2020-05-19 | 北京嘀嘀无限科技发展有限公司 | Text processing method and device, electronic equipment and storage medium |
| CN111178073B (en) * | 2018-10-23 | 2024-06-04 | 北京嘀嘀无限科技发展有限公司 | Text processing method, device, electronic equipment and storage medium |
| CN109543153A (en) * | 2018-11-13 | 2019-03-29 | 成都数联铭品科技有限公司 | A kind of sequence labelling system and method |
| CN109543153B (en) * | 2018-11-13 | 2023-08-18 | 成都数联铭品科技有限公司 | Sequence labeling system and method |
| CN111368541B (en) * | 2018-12-06 | 2024-06-11 | 北京搜狗科技发展有限公司 | Named entity identification method and device |
| CN111368541A (en) * | 2018-12-06 | 2020-07-03 | 北京搜狗科技发展有限公司 | Named entity identification method and device |
| CN109791570A (en) * | 2018-12-13 | 2019-05-21 | 香港应用科技研究院有限公司 | Efficiently and accurately name entity recognition method and device |
| CN111382570B (en) * | 2018-12-28 | 2024-05-03 | 深圳市优必选科技有限公司 | Text entity recognition method, device, computer equipment and storage medium |
| CN111382570A (en) * | 2018-12-28 | 2020-07-07 | 深圳市优必选科技有限公司 | Text entity recognition method and device, computer equipment and storage medium |
| CN109815296A (en) * | 2018-12-29 | 2019-05-28 | 北京中科闻歌科技股份有限公司 | The personage's construction of knowledge base method, apparatus and storage medium of notarization document |
| CN111488737A (en) * | 2019-01-09 | 2020-08-04 | 阿里巴巴集团控股有限公司 | Text recognition method, device and equipment |
| CN111488737B (en) * | 2019-01-09 | 2023-04-14 | 阿里巴巴集团控股有限公司 | Text recognition method, device and equipment |
| CN109886270A (en) * | 2019-01-17 | 2019-06-14 | 大连理工大学 | A kind of case element recognition methods towards electronics folder notes text |
| CN109886270B (en) * | 2019-01-17 | 2022-03-01 | 大连理工大学 | Case element identification method for electronic file record text |
| WO2020215456A1 (en) * | 2019-04-26 | 2020-10-29 | 网宿科技股份有限公司 | Text labeling method and device based on teacher forcing |
| EP3748548A4 (en) * | 2019-04-26 | 2021-03-10 | Wangsu Science & Technology Co., Ltd. | Adversarial learning-based text annotation method and device |
| EP3751445A4 (en) * | 2019-04-26 | 2021-03-10 | Wangsu Science & Technology Co., Ltd. | Text labeling method and device based on teacher forcing |
| CN110489727B (en) * | 2019-07-12 | 2023-07-07 | 深圳追一科技有限公司 | Person name recognition method and related device |
| CN110489727A (en) * | 2019-07-12 | 2019-11-22 | 深圳追一科技有限公司 | Name recognition methods and relevant apparatus |
| CN110688467A (en) * | 2019-08-23 | 2020-01-14 | 北京百度网讯科技有限公司 | Named entity recognition method and device, computer equipment and storage medium |
| CN110569332A (en) * | 2019-09-09 | 2019-12-13 | 腾讯科技（深圳）有限公司 | Sentence feature extraction processing method and device |
| CN110569332B (en) * | 2019-09-09 | 2023-01-06 | 腾讯科技（深圳）有限公司 | Sentence feature extraction processing method and device |
| CN110750990B (en) * | 2019-09-17 | 2024-09-06 | 平安科技（深圳）有限公司 | Labeling method, system, device and storage medium for entity recognition corpus |
| CN110750990A (en) * | 2019-09-17 | 2020-02-04 | 平安科技（深圳）有限公司 | Entity identification corpus labeling method, system, device and storage medium |
| CN110750991A (en) * | 2019-09-18 | 2020-02-04 | 平安科技（深圳）有限公司 | Entity identification method, device, equipment and computer readable storage medium |
| CN110750991B (en) * | 2019-09-18 | 2022-04-15 | 平安科技（深圳）有限公司 | Entity identification method, device, equipment and computer readable storage medium |
| CN111178075A (en) * | 2019-12-19 | 2020-05-19 | 厦门快商通科技股份有限公司 | Online customer service log analysis method, device and equipment |
| CN111125438A (en) * | 2019-12-25 | 2020-05-08 | 北京百度网讯科技有限公司 | Entity information extraction method and device, electronic equipment and storage medium |
| CN111125438B (en) * | 2019-12-25 | 2023-06-27 | 北京百度网讯科技有限公司 | Entity information extraction method and device, electronic equipment and storage medium |
| CN113051918B (en) * | 2019-12-26 | 2024-05-14 | 北京中科闻歌科技股份有限公司 | Named entity recognition method, device, equipment and medium based on ensemble learning |
| CN113051918A (en) * | 2019-12-26 | 2021-06-29 | 北京中科闻歌科技股份有限公司 | Named entity identification method, device, equipment and medium based on ensemble learning |
| CN111400429B (en) * | 2020-03-09 | 2023-06-30 | 北京奇艺世纪科技有限公司 | Text entry searching method, device, system and storage medium |
| CN111400429A (en) * | 2020-03-09 | 2020-07-10 | 北京奇艺世纪科技有限公司 | Text entry searching method, device, system and storage medium |
| CN111797629A (en) * | 2020-06-23 | 2020-10-20 | 平安医疗健康管理股份有限公司 | Medical text data processing method and device, computer equipment and storage medium |
| CN111797629B (en) * | 2020-06-23 | 2022-07-29 | 平安医疗健康管理股份有限公司 | Method and device for processing medical text data, computer equipment and storage medium |
| CN112270173A (en) * | 2020-10-27 | 2021-01-26 | 北京百度网讯科技有限公司 | Character mining method and device in text, electronic equipment and storage medium |
| CN112270173B (en) * | 2020-10-27 | 2021-10-26 | 北京百度网讯科技有限公司 | Character mining method and device in text, electronic equipment and storage medium |
| CN112541065A (en) * | 2020-12-11 | 2021-03-23 | 浙江汉德瑞智能科技有限公司 | Medical new word discovery processing method based on representation learning |
| CN113127645A (en) * | 2021-04-09 | 2021-07-16 | 厦门渊亭信息科技有限公司 | Automatic extraction method of large-scale knowledge graph ontology, terminal equipment and storage medium |
| CN113127060A (en) * | 2021-04-09 | 2021-07-16 | 中通服软件科技有限公司 | Software function point identification method based on natural language pre-training model (BERT) |
| CN113971216A (en) * | 2021-10-22 | 2022-01-25 | 北京百度网讯科技有限公司 | Data processing method and device, electronic equipment and memory |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107330011B (en) | 2019-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107330011B (en) | The recognition methods of the name entity of more strategy fusions and device | |
| CN110096570B (en) | Intention identification method and device applied to intelligent customer service robot | |
| CN110097085B (en) | Lyric text generation method, training method, device, server and storage medium | |
| CN106919673B (en) | Text mood analysis system based on deep learning | |
| US11113323B2 (en) | Answer selection using a compare-aggregate model with language model and condensed similarity information from latent clustering | |
| CN109960800A (en) | Weakly supervised text classification method and device based on active learning | |
| CN104050160B (en) | Interpreter's method and apparatus that a kind of machine is blended with human translation | |
| CN109635108B (en) | Man-machine interaction based remote supervision entity relationship extraction method | |
| CN113505200B (en) | Sentence-level Chinese event detection method combined with document key information | |
| CN110457689B (en) | Semantic processing method and related device | |
| CN108647225A (en) | A kind of electric business grey black production public sentiment automatic mining method and system | |
| CN108563703A (en) | A kind of determination method of charge, device and computer equipment, storage medium | |
| CN107122349A (en) | A kind of feature word of text extracting method based on word2vec LDA models | |
| CN106570180A (en) | Artificial intelligence based voice searching method and device | |
| CN115357719B (en) | Power audit text classification method and device based on improved BERT model | |
| CN114818717B (en) | Chinese named entity recognition method and system integrating vocabulary and syntax information | |
| CN110321561A (en) | A kind of keyword extracting method and device | |
| CN113360582B (en) | Relation classification method and system based on BERT model fusion multi-entity information | |
| CN107357765A (en) | Word document flaking method and device | |
| CN105988978B (en) | Determine the method and system of text focus | |
| CN104317882A (en) | Decision-based Chinese word segmentation and fusion method | |
| CN115714002B (en) | Training method for depression risk detection model, depression symptom early warning method and related equipment | |
| CN106933802A (en) | A kind of social security class entity recognition method and device towards multi-data source | |
| CN110610003A (en) | Method and system for assisting text annotation | |
| CN112036186A (en) | Corpus labeling method and device, computer storage medium and electronic equipment |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhao Honghong Inventor after: Wang Mengmeng Inventor after: Jin Yaohong Inventor after: Jiang Hongfei Inventor after: Yang Kaicheng Inventor after: Dong Mingtao Inventor before: Zhao Honghong Inventor before: Wang Mengmeng Inventor before: Jin Yaohong Inventor before: Jiang Hongfei Inventor before: Yang Kaicheng |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20190904 Address after: Room 630, 6th floor, Block A, Wanliu Xingui Building, 28 Wanquanzhuang Road, Haidian District, Beijing Patentee after: China Science and Technology (Beijing) Co., Ltd. Address before: Room 601, Block A, Wanliu Xingui Building, 28 Wanquanzhuang Road, Haidian District, Beijing Co-patentee before: China Science and Technology (Beijing) Co., Ltd. Patentee before: Beijing Shenzhou Taiyue Software Co., Ltd. |
|
| TR01 | Transfer of patent right | ||
| CP03 | Change of name, title or address |
Address after: 230000 zone B, 19th floor, building A1, 3333 Xiyou Road, hi tech Zone, Hefei City, Anhui Province Patentee after: Dingfu Intelligent Technology Co., Ltd Address before: Room 630, 6th floor, Block A, Wanliu Xingui Building, 28 Wanquanzhuang Road, Haidian District, Beijing Patentee before: DINFO (BEIJING) SCIENCE DEVELOPMENT Co.,Ltd. |
|
| CP03 | Change of name, title or address |