CN109344401A - Named Entity Extraction Model training method, name entity recognition method and device - Google Patents
Named Entity Extraction Model training method, name entity recognition method and device Download PDFInfo
- Publication number
- CN109344401A CN109344401A CN201811089249.9A CN201811089249A CN109344401A CN 109344401 A CN109344401 A CN 109344401A CN 201811089249 A CN201811089249 A CN 201811089249A CN 109344401 A CN109344401 A CN 109344401A
- Authority
- CN
- China
- Prior art keywords
- corpus information
- adjective
- identified
- word
- information
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the present application discloses a kind of Named Entity Extraction Model training method and name entity recognition method, for identifying corpus information to be identified according to object naming entity recognition model, to obtain the corresponding name entity of corpus information to be identified, the accuracy rate for identifying corpus information to be identified is improved.The embodiment of the present application method includes: to obtain sample corpus information, and the sample corpus information includes adverbial word corpus information and adjective corpus information；The target adverbial word corpus information for meeting presetting rule is determined from the adverbial word corpus information；Entity recognition is named to the adjective corpus information based on condition random field algorithm, obtains target adjective corpus information；According to the target adverbial word corpus information and the initial Named Entity Extraction Model of target adjective corpus information training, object naming entity recognition model is obtained, the object naming entity recognition model names entity for identification.
Description
Technical field
The invention relates to natural language processing field more particularly to a kind of Named Entity Extraction Model training sides
Method, name entity recognition method and device.
Background technique
The rapid development of internet at present drives the innovative change of many traditional industries, constantly by traditional industries and people
Work intelligently combines, and car networking constantly derives many people as swift and violent one of system is wherein developed, in vehicle diagnosis industry
Work intelligence blank, wherein the technology of core the most is then natural language processing in these derivatives, and Entity recognition is named to exist
It plays important role in the treatment process of natural language, while being also a very challenging property in Chinese text processing
Problem.
Existing name entity recognition techniques are mainly based upon the mode of statistics effectively to be extracted to key message, so
Entity recognition is named to the key message extracted afterwards.
In existing symptom of vehicle failure vocabulary, the syntactic structure that do not fix, same symptom of vehicle failure can be with
The combination of multiple vocabulary is resolved into, while different spare and accessory parts correspond to identical phenomenon of the failure.Therefore by way of statistics pair
Phenomenon of the failure, which is named Entity recognition, will appear the ambiguity due to word and is matched to the phenomenon of the failure of mistake or certain is existing
As frequency of occurrence is more in a certain components for vocabulary, the components are classified as once occurring, cause very big identification error, from
And it can not identify the vocabulary of symptom of vehicle failure.
Summary of the invention
The embodiment of the present application provides a kind of Named Entity Extraction Model training method and name entity recognition method, uses
In identifying corpus information to be identified according to object naming entity recognition model, to obtain the corresponding name of corpus information to be identified
Entity improves the accuracy rate for identifying corpus information to be identified.
The embodiment of the present application first aspect provides a kind of Named Entity Extraction Model training method, comprising:
Named Entity Extraction Model training device obtains sample corpus information, and the sample corpus information includes adverbial word corpus
Information and adjective corpus information；Named Entity Extraction Model training device meets preset rule from adverbial word corpus information determination
Target adverbial word corpus information then；Named Entity Extraction Model training device is based on condition random field algorithm and describes word to described
Material information is named Entity recognition, obtains target adjective corpus information；Named Entity Extraction Model training device is according to institute
Target adverbial word corpus information and the initial Named Entity Extraction Model of target adjective corpus information training are stated, target life is obtained
Name entity recognition model, the object naming entity recognition model name entity for identification.
Based on the embodiment of the present application in a first aspect, in the first implementation of the embodiment of the present application first aspect, name
Before entity recognition model training device determines the target adverbial word corpus information for meeting presetting rule from the adverbial word corpus information,
The method also includes: the sample corpus information is carried out Chinese word segmentation by Named Entity Extraction Model training device, obtains institute
State adverbial word corpus information and the adjective corpus information.
The first implementation based on the embodiment of the present application first aspect and first aspect, the embodiment of the present application first
In second of implementation of aspect, the Named Entity Extraction Model training device is based on condition random field algorithm to the shape
Hold word corpus information and be named Entity recognition, obtaining target adjective corpus information includes: Named Entity Extraction Model training
Device is respectively marked each word in the adjective corpus information using B, I and O label symbol, wherein the B is used
In the word for marking the adjective corpus information prefix, the I is used to mark the word in the adjective corpus information word, described
It is in corpus information except the word in the prefix and institute's predicate that O, which is used to mark described describe,；Named Entity Extraction Model training
Device is using the adjective corpus information construction feature collection of functions after label, and the characteristic function is for determining that the target is described
Word corpus information, the characteristic function is as shown in the first formula；
Wherein, i indicates the moment, and X is current observation sequence, yi-1For the first annotated sequence, yiFor the second annotated sequence,
Two annotated sequences are the corresponding annotated sequence of current observation sequence X, and the first annotated sequence is the previous mark of the second annotated sequence
Sequence, b (x, i) are the i-th moment real observation value, and NB is the fixation value of the first annotated sequence, and NI is the second annotated sequence
Fixed value, otherwise are expressed as except yi-1=NB, yiOutside=NI, f (yi-1, yi, x, i) and it is O.
The first implementation based on the embodiment of the present application first aspect and first aspect is to the second of first aspect
Any one of kind implementation, in the third implementation of the embodiment of the present application first aspect, Named Entity Extraction Model training
Device calculates the weight λ, the weight λ of the feature function set for training the target to order by maximum- likelihood estimation
Name entity recognition model.
The embodiment of the present application second aspect provides a kind of name entity recognition method, comprising:
Entity recognition device is named to obtain corpus information to be identified；Entity recognition device is named to believe the corpus to be identified
Breath carries out Chinese word segmentation, obtains adjective corpus information to be identified and adverbial word corpus information to be identified；Name Entity recognition dress
It sets and the adjective corpus information to be identified and the adverbial word corpus letter to be identified is identified according to object naming entity recognition model
Breath, respectively obtains the corresponding name entity of the adjective corpus information to be identified and the adverbial word corpus information to be identified is corresponding
Name entity, the object naming entity recognition model pass through target adverbial word corpus information and target adjective corpus information
Training obtains.
It is described in the first implementation of the embodiment of the present application second aspect based on the embodiment of the present application second aspect
Method further include: name entity recognition device according to the corresponding name entity of the adjective corpus information to be identified and it is described to
The corresponding name entity of identification adverbial word corpus information determine the recall rate of the object naming entity recognition model, rate of precision and
F value.
The first implementation based on the embodiment of the present application second aspect and second aspect, the embodiment of the present application second
In second of implementation of aspect, name entity recognition device obtains feedback information；Name entity recognition device according to
It feeds back confidence and corrects the object naming entity recognition model.
The embodiment of the present application third aspect provides a kind of Named Entity Extraction Model training device, the name Entity recognition
Model training apparatus includes: processor；And the memory being connect with the processor communication；Wherein, the memory storage
There is readable instruction, the readable instruction realizes such as first aspect when being executed by the Named Entity Extraction Model training device
Or the Message Processing or control operation of any possible implementation of first aspect.
A kind of name entity recognition device, the name entity recognition device packet are ordered in the offer of the embodiment of the present application fourth aspect
It includes: processor；And the memory being connect with the processor communication；Wherein, the memory is stored with readable instruction,
The readable instruction is realized when being executed by the name entity recognition device as second aspect or second aspect are any possible
The Message Processing or control operation of implementation.
The 5th aspect of the embodiment of the present application provides a kind of computer storage medium, is stored thereon with computer program, when
When the computer program is run on computers so that the computer execute above-mentioned first aspect or first aspect is any can
The method of the implementation of energy.
The 6th aspect of the embodiment of the present application provides a kind of computer storage medium, is stored thereon with computer program, when
When the computer program is run on computers so that the computer execute above-mentioned second aspect or second aspect is any can
The method of the implementation of energy.
The 7th aspect of the embodiment of the present application provides a kind of computer program product, which includes calculating
Machine software instruction, the computer software instructions can be loaded to realize that above-mentioned first aspect or first aspect are appointed by processor
The method of one possible implementation.
The embodiment of the present application eighth aspect provides a kind of computer program product, which includes calculating
Machine software instruction, the computer software instructions can be loaded to realize that above-mentioned second aspect or second aspect are appointed by processor
The method of one possible implementation.
As can be seen from the above technical solutions, the embodiment of the present application has the advantage that
Adverbial word corpus letter in the present embodiment, in the rule-based identification corpus information of Named Entity Extraction Model training device
Breath obtains target adverbial word corpus information, and identifies the adjective corpus information in corpus information based on condition random field algorithm,
Target adjective corpus information is obtained, to order based on target adverbial word corpus information and target adjective corpus information training objective
Name entity recognition model improves the efficiency of object naming entity recognition model identification name entity.
Detailed description of the invention
Fig. 1 is an application scenarios schematic diagram provided by the embodiments of the present application；
Fig. 2 is a kind of schematic flow chart of Named Entity Extraction Model training method provided by the embodiments of the present application；
Fig. 3 is a kind of schematic flow chart for naming entity recognition method provided by the embodiments of the present application；
Fig. 4 is a kind of schematic block diagram of Named Entity Extraction Model training device provided by the embodiments of the present application；
Fig. 5 is a kind of schematic block diagram for naming entity recognition device provided by the embodiments of the present application；
Fig. 6 is a kind of hardware structural diagram of Named Entity Extraction Model training device provided by the embodiments of the present application；
Fig. 7 is a kind of hardware structural diagram for naming entity recognition device provided by the embodiments of the present application.
Specific embodiment
With reference to the accompanying drawing, the technical solution in the application is described, it is clear that described embodiment is only this
Apply for the embodiment of a part, instead of all the embodiments.Those of ordinary skill in the art are it is found that going out with new technology
Existing, technical solution provided by the embodiments of the present application is equally applicable for similar technical problem.
The embodiment of the present application provides a kind of Named Entity Extraction Model training method and name entity recognition method, uses
In identifying corpus information to be identified according to object naming entity recognition model, to obtain the corresponding name of corpus information to be identified
Entity improves the accuracy rate for identifying corpus information to be identified.
The description and claims of this application and term " first " in above-mentioned attached drawing, " second " etc. are for distinguishing
Similar object, without being used to describe a particular order or precedence order.It should be understood that the data used in this way are in appropriate feelings
It can be interchanged under condition, so that the embodiments described herein can be real with the sequence other than the content for illustrating or describing herein
It applies.In addition, term " includes " and " having " and their any deformation, it is intended that cover it is non-exclusive include, for example, packet
The process, method, system, product or equipment for having contained series of steps or module those of be not necessarily limited to be clearly listed step or
Module, but may include other steps being not clearly listed or intrinsic for these process, methods, product or equipment or
Module.
Referring to FIG. 1, Fig. 1 is a kind of application scenarios schematic diagram provided by the embodiments of the present application, the application scenarios schematic diagram
Including corpus information to be identified, name entity recognition device and name entity.Wherein, real entity recognition device is named to be provided with
Object naming entity recognition model.It names entity recognition device to obtain corpus information to be identified, then passes through object naming entity
Identification model identifies corpus information to be identified, obtains the corresponding name entity of corpus information to be identified.
It should be noted that Named Entity Extraction Model training method provided by the embodiments of the present application and name entity are known
Other method can be applied to the identification of vehicle failure vocabulary, can be used for automobile intelligent voice dialogue, can be wide with user's automobile
The jettison system of announcement, the embodiment of the present application and subsequent embodiment are only retouched with the identification of vehicle failure vocabulary as an example
It states.
The application scenarios of the embodiment of the present application are described above, below from Named Entity Extraction Model training method to this Shen
Please embodiment be described.
Referring to FIG. 2, Fig. 2 is a kind of the schematic of Named Entity Extraction Model training method provided by the embodiments of the present application
Flow chart, as shown in Fig. 2, a kind of Named Entity Extraction Model training method provided by the embodiments of the present application may include following step
It is rapid:
201, Named Entity Extraction Model training device obtains sample corpus information.
Named Entity Extraction Model training device can obtain sample corpus from the database of symptom of vehicle failure vocabulary
Information.The sample corpus information includes adverbial word sample corpus information and adjective sample corpus information.For example, the sample corpus
Information can be " automobile air-conditioning refrigeration effect is bad ", wherein " bad " is adverbial word corpus information, " automobile air-conditioning refrigeration effect "
For adjective sample corpus information.
It should be noted that in the embodiment of the present application sample corpus information can be it is multiple, the application implement and it is subsequent
Embodiment is only described using " automobile air-conditioning refrigeration effect is bad " as sample corpus information.
202, Named Entity Extraction Model training device carries out Chinese word segmentation to sample corpus information.
Named Entity Extraction Model training device carries out Chinese word segmentation to the sample corpus information got, obtains adverbial word sample
This corpus information and adjective sample corpus information.For example, sample corpus information " automobile air-conditioning refrigeration effect is bad " is carried out
Participle, obtains " bad " and " automobile air-conditioning refrigeration effect ".Wherein, " bad " is adverbial word corpus information, " automobile air-conditioning refrigeration effect
Fruit " is adjective sample corpus information.
203, Named Entity Extraction Model training device determines the target adverbial word for meeting presetting rule from adverbial word corpus information
Corpus information.
Named Entity Extraction Model training device is after getting adverbial word corpus information in sample corpus information, by adverbial word
Corpus information is matched with the adverbial word rule in database, and the target adverbial word for meeting presetting rule is determined from adverbial word corpus information
Corpus information, and the target adverbial word corpus information for meeting presetting rule is indicated, then target adverbial word corpus information is turned
It moves in condition random field (conditional random fields, CRF) statistical module.
Specifically, in adverbial word rule, by symptom of vehicle failure vocabulary according to inverse document frequency (inverse document
Frequency, IDF) incremental form is ranked up, it is to be understood that symptom of vehicle failure vocabulary is arranged according to IDF
After sequence, when adverbial word corpus information is compared with adverbial word rule, it is retrieved in the symptom of vehicle failure vocabulary of sequence earlier above
Probability it is bigger.For example, the adverbial word corpus information that Named Entity Extraction Model training device is got is " bad " and adverbial word
Symptom of vehicle failure vocabulary in rule is matched, if determining automobile corresponding with adverbial word corpus information from adverbial word rule
Phenomenon of the failure vocabulary is " fine " and " very bad ", Named Entity Extraction Model training device judgement " fine " and " very not
It is good " whether meet presetting rule, which can be the similarity of adverbial word corpus information Yu symptom of vehicle failure vocabulary,
In the similarity can be set to 70%, it can be seen that vocabulary " very bad " and meeting for adverbial word corpus information " bad ", are preset
Therefore rule is transferred to CRF system by vocabulary " very bad " setting flag information, and by symptom of vehicle failure vocabulary " very bad "
Count module.
204, Named Entity Extraction Model training device orders adjective corpus information based on condition random field algorithm
Name Entity recognition.
Named Entity Extraction Model training device respectively marks adjective corpus information using B, I and O label symbol
Note, wherein B is used to mark the word of adjective corpus information prefix, and I is used to mark the word in adjective corpus information word, and O is used for
Label describes to be in corpus information except the word in prefix and word.
Specifically, Named Entity Extraction Model training device splits adjective corpus information, such as by adjective
Corpus information " automobile air-conditioning refrigeration effect " is split as " vapour, vehicle is empty, adjusts, and system is cold, effect, fruit ", then list entries x=vapour,
Vehicle, it is empty, adjust, system is cold, effect, fruit, the corresponding flag sequence of list entries is y={ B-SYS, I-SYS, B-SYS, I-SYS, B-
ADJ, I-ADJ, O, O }, wherein SYS indicates that system name, ADJ indicate adjective.
It is understood that Named Entity Extraction Model training device can get K shape from sample corpus information
Hold word corpus information, which is { x (k), y (k) }, and wherein k is adjective corpus
The number of information.
Named Entity Extraction Model training device uses the adjective corpus information construction feature collection of functions after label, the spy
Collection of functions is levied as shown in formula 2-1.
Wherein, i indicates the moment, and X is current observation sequence, yi-1For the first annotated sequence, yiFor the second annotated sequence,
Two annotated sequences are the corresponding annotated sequence of current observation sequence X, and the first annotated sequence is the previous mark of the second annotated sequence
Sequence, b (x, i) are the i-th moment real observation value, and NB is the fixation value of the first annotated sequence, and NI is the second annotated sequence
Fixed value, otherwise are expressed as except yi-1=NB, yiOutside=NI, f (yi-1, yi, x, i) and it is O.
For example, for adjective corpus information " automobile air-conditioning refrigeration effect ", i=1 moment corresponding ' vapour ' word, the i=2 moment
Corresponding ' vehicle ' word, i=3 moment corresponding ' sky ' word, i=4 moment corresponding ' tune ' word, i=5 moment corresponding ' system ' word, i=6 moment
Corresponding ' cold ' word, i=7 moment corresponding ' effect ' word, i=8 moment corresponding ' fruit ' word.In construction feature function, adjective is considered
The previous word of corpus information only considers the influence at i=3 moment that is, at the i=4 moment, without considering i=1,2,4,5,5,6,7,
8.For for the i=4 of adjective corpus information " automobile air-conditioning refrigeration effect ", x=tune, y=I-SYS, x-1=are empty, y-1
=B-SYS, if previous observation sequence is ' sky ', previous annotated sequence is B-SYS, and current mark series is I-SYS, at this time feature
Function representation is f=1.Otherwise, f=O.
The target adjective corpus information that Named Entity Extraction Model training device is obtained by the first formula, the target shape
Hold word corpus information available data sets T=(xk, yk) indicate, K is the quantity of data set.
205, Named Entity Extraction Model training device calculates weight to feature function set by maximum- likelihood estimation
λ。
Named Entity Extraction Model training device calculates respective weights λ, example to feature function set by maximal possibility estimation
Such as, the data set of target adjective corpus information is T=(xk, yk), data set probability is P (T | λ) under λ, data set T with combine
It is known to experience distribution P (x, y) and mutually indepedent between data, then train similarity function can be as shown in formula 2-2:
L (λ)=ΠX, yP (y | x, λ)P (x, y) (2-2)
Wherein:It is distributed for the experience of data set T.
Named Entity Extraction Model training device is carried out by estimating formula 2-3 derivation parameter, and using L-BFGS
Training.
Wherein, the desired value of Ep=distribution p, j-th of Named Entity Extraction Model of j=.
206, Named Entity Extraction Model training device is according to target adverbial word corpus information and target adjective corpus information
The initial Named Entity Extraction Model of training, obtains object naming entity recognition model.
It should be noted that name entity is known when marking the corresponding data set of adjective corpus information lower than preset threshold
Other model training apparatus introduces smoothing factorPass through input feature vector collection of functions and smooth factor pair object naming Entity recognition mould
Type is trained, and obtains corresponding feature function set and its weight.
Adverbial word corpus letter in the present embodiment, in the rule-based identification corpus information of Named Entity Extraction Model training device
Breath obtains target adverbial word corpus information, and identifies the adjective corpus information in corpus information based on condition random field algorithm,
Target adjective corpus information is obtained, to order based on target adverbial word corpus information and target adjective corpus information training objective
Name entity recognition model improves the efficiency of object naming entity recognition model identification name entity.
The embodiment of the present application is described from Named Entity Extraction Model training method above, is known below from name entity
The embodiment of the present application is described in other method.
Referring to FIG. 3, Fig. 3 is a kind of name entity recognition method schematic flow chart provided by the embodiments of the present application, such as
Shown in Fig. 3, a kind of name entity recognition method provided by the embodiments of the present application be may include steps of:
301, name entity recognition device obtains corpus information to be identified.
Name entity recognition device can obtain corpus information to be identified by automobile detection apparatus, the corpus letter to be identified
It ceases associated with symptom of vehicle failure vocabulary.
302, corpus information to be identified is carried out Chinese word segmentation by name entity recognition device.
Name entity recognition device gets corpus information to be identified and corpus information to be identified is carried out Chinese word segmentation later,
Obtain adjective corpus information to be identified and adverbial word corpus information to be identified.
303, name entity recognition device identifies the adjective corpus to be identified according to object naming entity recognition model
Information and the adverbial word corpus information to be identified.
Name entity recognition device according to object naming entity recognition model identify adjective corpus information to be identified and to
It identifies adverbial word corpus information, respectively obtains the corresponding name entity of adjective corpus information to be identified and adverbial word corpus to be identified letter
Corresponding name entity is ceased, object naming entity recognition model is believed by target adverbial word corpus information and target adjective corpus
Breath training obtains.
304, name entity recognition device is according to the corresponding name entity of the adjective corpus information to be identified and described
The corresponding name entity of adverbial word corpus information to be identified determine the recall rate of the object naming entity recognition model, rate of precision with
And F value.
Name entity recognition device determines the recall rate, accurate of the object naming entity recognition model by confusion matrix
Rate and F value.Specifically, which can be as shown in table 3-1.
3-1. confusion matrix
| The adjectival quantity of identification | The non-adjective quantity of identification | |
| Practical adjective quantity Z | TP | FN |
| Practical non-adjective quantity P | FP | TN |
Wherein, recall rate: Recall=TP/ (TP+FN),
Rate of precision: Precision=TP/ (TP+FP),
F value: F1=2*Recall*Precision/ (Recall+Precision).
Name entity recognition device determines describing for identification from the corresponding name entity of adjective corpus information to be identified
The quantity TP of word and non-adjective quantity FN.
Entity recognition device is named to name the shape for determining and identifying in entity from non-adjective corpus information to be identified is corresponding
Hold the quantity FP and non-adjective quantity TN of word.
It names entity recognition device to determine recall rate by TP and FN, rate of precision is determined by TP and FP, by calling together
After the rate of returning and rate of precision determine F value, recall rate, rate of precision and F value are compared with preset threshold value respectively, if recalling
Rate, rate of precision and/or F value are less than preset threshold value, then name entity recognition device update or re -training object naming entity
Identification model.
Optionally, in the present embodiment, name entity recognition device can also repair object naming entity recognition model
Just.
Specifically, when user by name entity recognition device get the corresponding name entity of corpus information to be identified it
Afterwards, the phenomenon of the failure of automobile is determined according to the name Entity recognition, if the name entity and the phenomenon of the failure of automobile be not corresponding,
User can be in name entity recognition device input feedback information, for example, the life that user gets from name entity recognition device
Name entity is " air-conditioning does not freeze ", and user it is actually detected to the phenomenon of the failure of automobile be " air-conditioning does not heat ", then user
" air-conditioning does not heat " can be fed back to name entity recognition device, name entity recognition device can be according to " air-conditioning heat "
Object naming entity recognition model is revised, so as to improve object naming entity recognition model identification name entity
Accuracy rate.
In the present embodiment, name entity recognition device identifies adverbial word corpus to be identified by object naming entity recognition model
Information and adjective corpus information to be identified obtain the corresponding name entity of adverbial word corpus information to be identified and shape to be identified
Hold the corresponding name Entity recognition of word corpus information, wherein the object naming entity recognition model is by sample corpus trained
The model arrived.Therefore, the present embodiment identifies corpus information to be identified by object naming entity recognition model, and it is real to improve name
The accuracy rate of body identification.
The embodiment of the present application is described from method above, below from the angle of device to involved by the embodiment of the present application
And device be described.
Referring to FIG. 4, Fig. 4 is a kind of the schematic of Named Entity Extraction Model training device provided by the embodiments of the present application
Block diagram.Wherein, Named Entity Extraction Model training device includes:
Acquiring unit 401, for obtaining sample corpus information, the sample corpus information includes adverbial word corpus information and shape
Hold word corpus information；
Determination unit 402, for determining the target adverbial word corpus information for meeting presetting rule from the adverbial word corpus information；
Recognition unit 403 is known for being named entity to the adjective corpus information based on condition random field algorithm
Not, target adjective corpus information is obtained；
Training unit 404, for according to the target adverbial word corpus information and target adjective corpus information training
Initial Named Entity Extraction Model obtains object naming entity recognition model, and the object naming entity recognition model is for knowing
It Ming Ming not entity.
Optionally, in the present embodiment, Named Entity Extraction Model training device further include:
Participle unit 405, for by the sample corpus information carry out Chinese word segmentation, obtain the adverbial word corpus information and
The adjective corpus information.
Optionally, in the present embodiment, recognition unit 403 is specifically used for:
Each word in the adjective corpus information is marked using B, I and O label symbol respectively, wherein institute
B is stated for marking the word of the adjective corpus information prefix, the I is for marking in the adjective corpus information word
Word, it is in corpus information except the word in the prefix and institute's predicate that the O, which is used to mark described describe,；
Using the adjective corpus information construction feature collection of functions after label, the characteristic function is for determining the target
Adjective corpus information, the characteristic function is as shown in the first formula；
Wherein, i indicates the moment, and X is current observation sequence, yi-1For the first annotated sequence, yiFor the second annotated sequence,
Two annotated sequences are the corresponding annotated sequence of current observation sequence X, and the first annotated sequence is the previous mark of the second annotated sequence
Sequence, b (x, i) are the i-th moment real observation value, and NB is the fixation value of the first annotated sequence, and NI is the second annotated sequence
Fixed value, otherwise are expressed as except yi-1=NB, yiOutside=NI, f (yi-1, yi, x, i) and it is 0.
Optionally, in the present embodiment, Named Entity Extraction Model training device further include:
Computing unit 406, for calculating the weight λ of the feature function set, the power by maximum- likelihood estimation
Weight λ is for training the object naming entity recognition model.
In the present embodiment, the rule-based adverbial word corpus information identified in corpus information of recognition unit 403 obtains target pair
Word corpus information, and based on the adjective corpus information in condition random field algorithm identification corpus information, obtain target and describe
Word corpus information, to name Entity recognition mould based on target adverbial word corpus information and target adjective corpus information training objective
Type improves the efficiency of object naming entity recognition model identification name entity.
Referring to FIG. 5, Fig. 5 is a kind of schematic block diagram for naming entity recognition device provided by the embodiments of the present application.Its
In, name entity recognition device includes:
First acquisition unit 501, for obtaining corpus information to be identified；
Participle unit 502 obtains adjective corpus to be identified for the corpus information to be identified to be carried out Chinese word segmentation
Information and adverbial word corpus information to be identified；
Recognition unit 503, for identifying the adjective corpus information to be identified according to object naming entity recognition model
With the adverbial word corpus information to be identified, the corresponding name entity of the adjective corpus information to be identified and described is respectively obtained
The corresponding name entity of adverbial word corpus information to be identified, the object naming entity recognition model pass through target adverbial word corpus information
And the training of target adjective corpus information obtains.
Optionally, in the present embodiment, entity recognition device is named further include:
Determination unit 504, for according to the corresponding name entity of the adjective corpus information to be identified and described wait know
Corresponding recall rate, rate of precision and the F for naming entity to determine the object naming entity recognition model of other adverbial word corpus information
Value.
Updating unit 505 is updated for being less than preset threshold when the recall rate, the rate of precision and/or the F value
The object naming identification model.
Optionally, in the present embodiment, entity recognition device is named further include:
Second acquisition unit 506, for obtaining feedback information；
Amending unit 507, for correcting the object naming entity recognition model according to the feedback confidence.
In the present embodiment, recognition unit 503 identifies adverbial word corpus information to be identified by object naming entity recognition model
And adjective corpus information to be identified, obtain the corresponding name entity of adverbial word corpus information to be identified and adjective to be identified
The corresponding name Entity recognition of corpus information, wherein the object naming entity recognition model is obtained by the training of sample corpus
Model.Therefore, the present embodiment identifies corpus information to be identified by object naming entity recognition model, improves name entity and knows
Other accuracy rate.
The embodiment of the present application also provides another Named Entity Extraction Model training device, referring to Fig. 6, Fig. 6 is the application
A kind of hardware structural diagram for Named Entity Extraction Model training device that embodiment provides, wherein name Entity recognition mould
Type training device includes: at least one processor 610, memory 650, transceiver 630 and bus system 620.
At least one described processor 610, the memory 650 and the transceiver 630 respectively with the bus system
620 are connected.
The transceiver 630 may include Receiver And Transmitter, the memory 650 may include read-only memory and/or with
Machine accesses memory, and provides operational order and data to processor 610.The a part of of memory 650 can also include non-easy
The property lost random access memory (NVRAM).
In some embodiments, memory 650 stores following element, executable modules or data structures, or
Their subset of person or their superset.
In the embodiment of the present application, by calling the operational order of the storage of memory 650, (operational order is storable in behaviour
Make in system), execute corresponding operation.Processor 610 controls the operation of Named Entity Extraction Model training device, processor
610 can also be known as CPU (Central Processing Unit, central processing unit).Memory 650 may include read-only
Memory and random access memory, and instruction and data is provided to processor 610.The a part of of memory 650 can also wrap
Include nonvolatile RAM (NVRAM).Each group of Named Entity Extraction Model training device in specific application
Part is coupled by bus system 620, and wherein bus system 620 can also include power supply in addition to including data/address bus
Bus, control bus and status signal bus in addition etc..But for the sake of clear explanation, various buses are all designated as bus in figure
System 620.
The method that above-mentioned the embodiment of the present application discloses can be applied in processor 610, or be realized by processor 610.
Processor 610 may be a kind of IC chip, the processing capacity with signal.During realization, the above method it is each
Step can be completed by the integrated logic circuit of the hardware in processor 610 or the instruction of software form.Above-mentioned processing
Device 610 can be general processor, digital signal processor (DSP), specific integrated circuit (ASIC), ready-made programmable gate array
(FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.May be implemented or
Person executes disclosed each method, step and logic diagram in the embodiment of the present application.General processor can be microprocessor or
Person's processor is also possible to any conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present application, can be straight
Connect and be presented as that hardware decoding processor executes completion, or in decoding processor hardware and software module combination executed
At.Software module can be located at random access memory, and flash memory, read-only memory, programmable read only memory or electrically-erasable can
In the storage medium of this fields such as programmable memory, register maturation.The storage medium is located at memory 650, the memory 650
It can be physically separate unit, be also possible to integrate with processor 610, processor 610 reads memory 650
In information, in conjunction with its hardware complete the above method the step of.
Wherein, the transceiver 630 includes adverbial word corpus letter for obtaining sample corpus information, the sample corpus information
Breath and adjective corpus information；
At least one described processor 610 is used to determine the target adverbial word for meeting presetting rule from the adverbial word corpus information
Corpus information；
At least one described processor 610 is also used to carry out the adjective corpus information based on condition random field algorithm
Entity recognition is named, target adjective corpus information is obtained；
At least one described processor 610 is also used to describe word according to the target adverbial word corpus information and the target
Expect the initial Named Entity Extraction Model of information training, obtains object naming entity recognition model, the object naming Entity recognition
Model names entity for identification.
In alternatively possible implementation, at least one described processor 610 is also used to the sample corpus information
Chinese word segmentation is carried out, the adverbial word corpus information and the adjective corpus information are obtained.
In alternatively possible implementation, at least one described processor 610 is specifically used for:
Each word in the adjective corpus information is marked using B, I and O label symbol respectively, wherein institute
B is stated for marking the word of the adjective corpus information prefix, the I is for marking in the adjective corpus information word
Word, it is word in corpus information in addition to the word in the prefix and institute's predicate that the O, which is used to marking described describe,；
Using the adjective corpus information construction feature collection of functions after label, the characteristic function is for determining the target
Adjective corpus information, the characteristic function is as shown in the first formula；
Wherein, i indicates the moment, and X is current observation sequence, yi-1For the first annotated sequence, yiFor the second annotated sequence,
Two annotated sequences are the corresponding annotated sequence of current observation sequence X, and the first annotated sequence is the previous mark of the second annotated sequence
Sequence, b (x, i) are the i-th moment real observation value, and NB is the fixation value of the first annotated sequence, and NI is the second annotated sequence
Fixed value, otherwise are expressed as except yi-1=NB, yiOutside=NI, f (yi-1, yi, x, i) and it is 0.
In a kind of possible implementation, at least one described processor 610 is specifically also used to: passing through maximal possibility estimation
Algorithm calculates the weight λ, the weight λ of the feature function set for training the object naming entity recognition model.
The embodiment of the present application also provides another name entity recognition device, referring to Fig. 7, Fig. 7 mentions for the embodiment of the present application
The hardware structural diagram of a kind of name entity recognition device supplied, wherein name entity recognition device includes: at least one
Manage device 710, memory 750, transceiver 730 and bus system 720.
At least one described processor 710, the memory 750 and the transceiver 730 respectively with the bus system
720 are connected.
In the present embodiment, at least one processor 710, memory 750 and transceiver 730 respectively correspond functional structure with before
State similar described in Fig. 6 corresponding embodiment, details are not described herein again.
Wherein, the transceiver 730 is for obtaining corpus information to be identified；
At least one described processor 710 is used to the corpus information to be identified carrying out Chinese word segmentation, obtains to be identified
Adjective corpus information and adverbial word corpus information to be identified；
At least one described processor 710 is also used to identify described to be identified describe according to object naming entity recognition model
Word corpus information and the adverbial word corpus information to be identified respectively obtain the corresponding name of the adjective corpus information to be identified
Entity and the corresponding name entity of the adverbial word corpus information to be identified, the object naming entity recognition model pass through target pair
Word corpus information and the training of target adjective corpus information obtain.
In a kind of possible implementation, at least one described processor 710 is also used to according to the adjective to be identified
The corresponding name entity of corpus information and the corresponding name entity of the adverbial word corpus information to be identified determine the object naming
Recall rate, rate of precision and the F value of entity recognition model；
If the recall rate, the rate of precision and/or the F value are less than preset threshold, update the object naming and know
Other model.
In alternatively possible implementation, the transceiver 730 is also used to obtain feedback information；
At least one described processor 710 is also used to correct the object naming Entity recognition mould according to the feedback confidence
Type.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided herein, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components
It can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown or
The mutual coupling, direct-coupling or communication connection discussed can be through some interfaces, the indirect coupling of device or unit
It closes or communicates to connect, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, each functional unit in each embodiment of the application can integrate in one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the application
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can store journey
The medium of sequence code.
The above, above embodiments are only to illustrate the technical solution of the application, rather than its limitations；Although referring to before
Embodiment is stated the application is described in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features；And these
It modifies or replaces, the spirit and scope of each embodiment technical solution of the application that it does not separate the essence of the corresponding technical solution.
Claims (11)
1. a kind of Named Entity Extraction Model training method characterized by comprising
Sample corpus information is obtained, the sample corpus information includes adverbial word corpus information and adjective corpus information；
The target adverbial word corpus information for meeting presetting rule is determined from the adverbial word corpus information；
Entity recognition is named to the adjective corpus information based on condition random field algorithm, obtains target adjective corpus
Information；
Initial Named Entity Extraction Model is trained according to the target adverbial word corpus information and the target adjective corpus information,
Object naming entity recognition model is obtained, the object naming entity recognition model names entity for identification.
2. meeting presetting rule from adverbial word corpus information determination according to the method described in claim 1, its feature is being
Target adverbial word corpus information before, the method also includes:
The sample corpus information is subjected to Chinese word segmentation, obtains the adverbial word corpus information and the adjective corpus information.
3. method according to claim 1 or 2, which is characterized in that described to be described based on condition random field algorithm to described
Word corpus information is named Entity recognition, obtains target adjective corpus information and includes:
Each word in the adjective corpus information is marked using B, I and O label symbol respectively, wherein the B is used
In the word for marking the adjective corpus information prefix, the I is used to mark the word in the adjective corpus information word, described
It is word in corpus information in addition to the word in the prefix and institute's predicate that O, which is used to marking described describe,；
Using the adjective corpus information construction feature collection of functions after label, the characteristic function is for determining that the target is described
Word corpus information, the characteristic function is as shown in the first formula；
Wherein, i indicates the moment, and X is current observation sequence, yi-1For the first annotated sequence, yiFor the second annotated sequence, the second mark
Sequence is the corresponding annotated sequence of current observation sequence X, and the first annotated sequence is the previous annotated sequence of the second annotated sequence, b
(x, i) is the i-th moment real observation value, and NB is the fixation value of the first annotated sequence, and NI is that the fixation of the second annotated sequence takes
Value, otherwise are expressed as except yi-1=NB, yiOutside=NI, f (yi-1, yi, x, i) and it is O.
4. according to the method described in claim 3, it is characterized in that, the method also includes:
The weight λ, the weight λ of the feature function set are calculated for training the target to order by maximum- likelihood estimation
Name entity recognition model.
5. a kind of name entity recognition method characterized by comprising
Obtain corpus information to be identified；
The corpus information to be identified is subjected to Chinese word segmentation, obtains adjective corpus information to be identified and adverbial word language to be identified
Expect information；
The adjective corpus information to be identified and the adverbial word corpus to be identified are identified according to object naming entity recognition model
Information respectively obtains the corresponding name entity of the adjective corpus information to be identified and the adverbial word corpus information pair to be identified
The name entity answered, the object naming entity recognition model are believed by target adverbial word corpus information and target adjective corpus
Breath training obtains.
6. according to the method described in claim 5, it is characterized in that, the method also includes:
It is corresponding according to the corresponding name entity of adjective corpus information to be identified and the adverbial word corpus information to be identified
Name entity determines the recall rate, rate of precision and F value of the object naming entity recognition model；
If the recall rate, the rate of precision and/or the F value are less than preset threshold, the object naming identification mould is updated
Type.
7. method according to claim 5 or 6, which is characterized in that the method also includes:
Obtain feedback information；
The object naming entity recognition model is corrected according to the feedback confidence.
8. a kind of Named Entity Extraction Model training device characterized by comprising
Acquiring unit, for obtaining sample corpus information, the sample corpus information includes adverbial word corpus information and describes word
Expect information；
Determination unit, for determining the target adverbial word corpus information for meeting presetting rule from the adverbial word corpus information；
Recognition unit is obtained for being named Entity recognition to the adjective corpus information based on condition random field algorithm
Target adjective corpus information；
Training unit, for according to the target adverbial word corpus information and the initial name of target adjective corpus information training
Entity recognition model, obtains object naming entity recognition model, and the object naming entity recognition model names reality for identification
Body.
9. a kind of name entity recognition device characterized by comprising
Acquiring unit, for obtaining corpus information to be identified；
Participle unit, for will the corpus information to be identified progress Chinese word segmentation, obtain adjective corpus information to be identified with
And adverbial word corpus information to be identified；
Recognition unit, for according to object naming entity recognition model identify the adjective corpus information to be identified and it is described to
It identifies adverbial word corpus information, respectively obtains the corresponding name entity of the adjective corpus information to be identified and the pair to be identified
The corresponding name entity of word corpus information, the object naming entity recognition model pass through target adverbial word corpus information and target
The training of adjective corpus information obtains.
10. a kind of server, which is characterized in that the server includes:
Processor；And
The memory being connect with the processor communication；
Wherein, the memory is stored with readable instruction, and the readable instruction is realized when being executed by the processor as weighed
Benefit require any one of 1 to 4 described in Named Entity Extraction Model training method.
11. a kind of server, which is characterized in that the server includes:
Processor；And
The memory being connect with the processor communication；
Wherein, the memory is stored with readable instruction, and the readable instruction is realized when being executed by the processor as weighed
Benefit require any one of 5 to 7 described in name entity recognition method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811089249.9A CN109344401B (en) | 2018-09-18 | 2018-09-18 | Named entity recognition model training method, named entity recognition method and named entity recognition device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811089249.9A CN109344401B (en) | 2018-09-18 | 2018-09-18 | Named entity recognition model training method, named entity recognition method and named entity recognition device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109344401A true CN109344401A (en) | 2019-02-15 |
| CN109344401B CN109344401B (en) | 2023-04-28 |
Family
ID=65306069
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811089249.9A Active CN109344401B (en) | 2018-09-18 | 2018-09-18 | Named entity recognition model training method, named entity recognition method and named entity recognition device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109344401B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110147551A (en) * | 2019-05-14 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Multi-class entity recognition model training, entity recognition method, server and terminal |
| CN110837737A (en) * | 2019-11-11 | 2020-02-25 | 中国电子科技集团公司信息科学研究院 | Method for recognizing ability word entity |
| CN111159535A (en) * | 2019-12-05 | 2020-05-15 | 北京声智科技有限公司 | Resource acquisition method and device |
| CN111222335A (en) * | 2019-11-27 | 2020-06-02 | 上海眼控科技股份有限公司 | Corpus correction method and device, computer equipment and computer-readable storage medium |
| CN112101026A (en) * | 2019-06-18 | 2020-12-18 | 掌阅科技股份有限公司 | Corpus sample set construction method, computing device and computer storage medium |
| CN112818689A (en) * | 2019-11-15 | 2021-05-18 | 马上消费金融股份有限公司 | Entity identification method, model training method and device |
| CN114548103A (en) * | 2020-11-25 | 2022-05-27 | 马上消费金融股份有限公司 | Training method of named entity recognition model and recognition method of named entity |
| CN112101026B (en) * | 2019-06-18 | 2024-05-31 | 掌阅科技股份有限公司 | Corpus sample set construction method, computing equipment and computer storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598510A (en) * | 2014-10-16 | 2015-05-06 | 苏州大学 | Event trigger word recognition method and device |
| WO2018023981A1 (en) * | 2016-08-03 | 2018-02-08 | 平安科技（深圳）有限公司 | Public opinion analysis method, device, apparatus and computer readable storage medium |
| CN107797994A (en) * | 2017-09-26 | 2018-03-13 | 昆明理工大学 | Vietnamese noun phrase block identifying method based on constraints random field |
| CN108255816A (en) * | 2018-03-12 | 2018-07-06 | 北京神州泰岳软件股份有限公司 | A kind of name entity recognition method, apparatus and system |
-
2018
- 2018-09-18 CN CN201811089249.9A patent/CN109344401B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598510A (en) * | 2014-10-16 | 2015-05-06 | 苏州大学 | Event trigger word recognition method and device |
| WO2018023981A1 (en) * | 2016-08-03 | 2018-02-08 | 平安科技（深圳）有限公司 | Public opinion analysis method, device, apparatus and computer readable storage medium |
| CN107797994A (en) * | 2017-09-26 | 2018-03-13 | 昆明理工大学 | Vietnamese noun phrase block identifying method based on constraints random field |
| CN108255816A (en) * | 2018-03-12 | 2018-07-06 | 北京神州泰岳软件股份有限公司 | A kind of name entity recognition method, apparatus and system |
Non-Patent Citations (1)
| Title |
|---|
| 于江德等: "三位一体字标注的汉语词法分析", 《中文信息学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110147551A (en) * | 2019-05-14 | 2019-08-20 | 腾讯科技（深圳）有限公司 | Multi-class entity recognition model training, entity recognition method, server and terminal |
| CN110147551B (en) * | 2019-05-14 | 2023-07-11 | 腾讯科技（深圳）有限公司 | Multi-category entity recognition model training, entity recognition method, server and terminal |
| CN112101026A (en) * | 2019-06-18 | 2020-12-18 | 掌阅科技股份有限公司 | Corpus sample set construction method, computing device and computer storage medium |
| CN112101026B (en) * | 2019-06-18 | 2024-05-31 | 掌阅科技股份有限公司 | Corpus sample set construction method, computing equipment and computer storage medium |
| CN110837737A (en) * | 2019-11-11 | 2020-02-25 | 中国电子科技集团公司信息科学研究院 | Method for recognizing ability word entity |
| CN112818689A (en) * | 2019-11-15 | 2021-05-18 | 马上消费金融股份有限公司 | Entity identification method, model training method and device |
| CN111222335A (en) * | 2019-11-27 | 2020-06-02 | 上海眼控科技股份有限公司 | Corpus correction method and device, computer equipment and computer-readable storage medium |
| CN111159535A (en) * | 2019-12-05 | 2020-05-15 | 北京声智科技有限公司 | Resource acquisition method and device |
| CN114548103A (en) * | 2020-11-25 | 2022-05-27 | 马上消费金融股份有限公司 | Training method of named entity recognition model and recognition method of named entity |
| CN114548103B (en) * | 2020-11-25 | 2024-03-29 | 马上消费金融股份有限公司 | Named entity recognition model training method and named entity recognition method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109344401B (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109344401A (en) | Named Entity Extraction Model training method, name entity recognition method and device | |
| CN107957993B (en) | English sentence similarity calculation method and device | |
| CN110457675A (en) | Prediction model training method, device, storage medium and computer equipment | |
| CN108845988B (en) | Entity identification method, device, equipment and computer readable storage medium | |
| CN111241814B (en) | Error correction method and device for voice recognition text, electronic equipment and storage medium | |
| CN111507419A (en) | Training method and device of image classification model | |
| CN110134969B (en) | Entity identification method and device | |
| WO2022121178A1 (en) | Training method and apparatus and recognition method and apparatus for text error correction model, and computer device | |
| CN111178537B (en) | Feature extraction model training method and device | |
| US20190026655A1 (en) | Machine Learning System for Patient Similarity | |
| CN113343703B (en) | Medical entity classification extraction method and device, electronic equipment and storage medium | |
| CN109523995B (en) | Audio recognition method, speech recognition equipment, readable storage medium storing program for executing and electronic equipment | |
| JP2018147392A (en) | Model learning device, score calculation device, method, data structure, and program | |
| CN111696661A (en) | Patient clustering model construction method, patient clustering method and related equipment | |
| CN108182182B (en) | Method and device for matching documents in translation database and computer readable storage medium | |
| CN110597956A (en) | Searching method, searching device and storage medium | |
| CN110489727B (en) | Person name recognition method and related device | |
| CN108877924A (en) | A kind of asthma method of determining probability and device | |
| CN116578704A (en) | Text emotion classification method, device, equipment and computer readable medium | |
| US20120221574A1 (en) | High-accuracy similarity search system | |
| US11948387B2 (en) | Optimized policy-based active learning for content detection | |
| CN111968740B (en) | Diagnostic label recommendation method and device, storage medium and electronic equipment | |
| CN110909157B (en) | Text classification method and device, computing equipment and readable storage medium | |
| US20200234134A1 (en) | Systems providing a learning controller utilizing indexed memory and methods thereto | |
| CN115631823A (en) | Similar case recommendation method and system |
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 |