CN106708806A - Sample determination method, device and system - Google Patents
Sample determination method, device and system Download PDFInfo
- Publication number
- CN106708806A CN106708806A CN201710031626.2A CN201710031626A CN106708806A CN 106708806 A CN106708806 A CN 106708806A CN 201710031626 A CN201710031626 A CN 201710031626A CN 106708806 A CN106708806 A CN 106708806A
- Authority
- CN
- China
- Prior art keywords
- sample
- confirmed
- similar
- confidence level
- similarity
- 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
- G06F40/205—Parsing
- G06F40/226—Validation
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 invention provides a sample determination method, device and system. The method comprises the steps of obtaining the credibility of each to-be-determined sample in a to-be-determined sample set; determining a similar sample of each to-be-determined sample in the to-be-determined sample set; correcting the credibility of each to-be-determined sample according to the similar sample, and obtaining the corrected credibility of each to-be-determined sample; determining each to-be-determined sample according to the corrected credibility of each to-be-determined sample. The method can improve the sample determination accuracy and then improve the application effect.
Description
Technical field
The application is related to natural language processing technique field, more particularly to a kind of sample confirmation method and apparatus and system.
Background technology
Increasingly mature with artificial intelligence technology, increasing application system has used artificial intelligence correlation technique,
Such as keyword retrieval, authentication, speaker separate, speaker's gender identity, be required for carrying out candidate samples final
Confirm, to confirm whether candidate samples are target samples, therefore, the degree of accuracy that sample confirms directly affects final application effect
Really.
In correlation technique, sample confirmation is directly typically carried out according to the confidence level of sample to be confirmed, when specifically confirming, setting
Confidence threshold value, judges whether the confidence level of sample to be confirmed exceedes threshold value set in advance, if it does, then sample confirms into
Work(;Otherwise, sample confirms failure.The above method only accounts for the information of single sample when sample confirms, and sample to be confirmed
Often disturbed by external factor such as environment or channels in actual applications, single sample information easily changes, if only examined
Consider the information of single sample, when sample confirms, can frequently result in sample to be confirmed and be identified mistake, i.e. target sample to be confirmed
Target sample is confirmed as by mistake by confirm as non-targeted samples or the non-targeted sample to be confirmed of mistake, sample is greatly reduced
The degree of accuracy of this confirmation, influences application effect.
The content of the invention
The application is intended at least solve to a certain extent one of technical problem in correlation technique.
Therefore, a purpose of the application is to propose a kind of sample confirmation method, the method can improve sample confirmation
The degree of accuracy, and then improve application effect.
Further object is to propose that a kind of sample confirms device.
Further object is to propose that a kind of sample confirms system.
To reach above-mentioned purpose, the sample confirmation method that the application first aspect embodiment is proposed, including:Obtain to be confirmed
The confidence level of each sample to be confirmed in sample set;Determine the similar sample of each sample to be confirmed in sample set to be confirmed
This;The confidence level of each sample to be confirmed is modified according to the similar sample, obtains the amendment of each sample to be confirmed
Confidence level afterwards;Revised confidence level according to each sample to be confirmed confirms to each sample to be confirmed.
To reach above-mentioned purpose, the sample that the application second aspect embodiment is proposed confirms device, including:Acquisition module,
Confidence level for obtaining each sample to be confirmed in sample set to be confirmed;Determining module, for determining sample set to be confirmed
The similar sample of each sample to be confirmed in conjunction;Correcting module, for according to the similar sample to each sample to be confirmed
Confidence level is modified, and obtains the revised confidence level of each sample to be confirmed;Module is confirmed, for be confirmed according to each
The revised confidence level of sample confirms to each sample to be confirmed.
To reach above-mentioned purpose, the sample that the application third aspect embodiment is proposed confirms system, including:Client, uses
In the sample to be confirmed for receiving user input;Service end, the sample to be confirmed for receiving client transmission, obtains sample to be confirmed
The confidence level of each sample to be confirmed in this set;Determine the similar sample of each sample to be confirmed in sample set to be confirmed;
The confidence level of each sample to be confirmed is modified according to the similar sample, obtains the revised of each sample to be confirmed
Confidence level;Revised confidence level according to each sample to be confirmed confirms to each sample to be confirmed.
The embodiment of the present application, by determining the similar sample of sample to be confirmed, and treats confirmatory sample according to similar sample
Confidence level be modified, and sample confirmation is carried out according to revised confidence level, can effectively lift the standard of sample confirmation
Exactness, especially corrects sample of the previous belief near confidence threshold value, and lifting effect becomes apparent from.
The aspect and advantage that the application is added will be set forth in part in the description, and will partly become from the following description
Substantially, or recognized by the practice of the application.
Brief description of the drawings
The above-mentioned and/or additional aspect of the application and advantage will become from the following description of the accompanying drawings of embodiments
Substantially and be readily appreciated that, wherein:
Fig. 1 is the schematic flow sheet of the sample confirmation method that the application one embodiment is proposed;
Fig. 2 is to carry out the right of sample confirmation respectively using confidence level after amendment previous belief and amendment in the embodiment of the present application
Compare schematic diagram;
Fig. 3 is the schematic flow sheet of the method for the similar sample that sample to be confirmed is determined in the embodiment of the present application;
Fig. 4 is the schematic flow sheet of the sample confirmation method of the application another embodiment proposition;
Fig. 5 is the structural representation that the sample that the application one embodiment is proposed confirms device;
Fig. 6 is the structural representation of the sample confirmation device of the application another embodiment proposition;
Fig. 7 is the structural representation that the sample that the application one embodiment is proposed confirms system.
Specific embodiment
Embodiments herein is described below in detail, the example of the embodiment is shown in the drawings, wherein from start to finish
Same or similar label represents same or similar module or the module with same or like function.Below with reference to attached
It is exemplary to scheme the embodiment of description, is only used for explaining the application, and it is not intended that limitation to the application.Conversely, this
The embodiment of application includes all changes fallen into the range of the spiritual and intension of attached claims, modification and is equal to
Thing.
Fig. 1 is the schematic flow sheet of the sample confirmation method that the application one embodiment is proposed.
As shown in figure 1, the method for the present embodiment includes:
S11:Obtain the confidence level of each sample to be confirmed in sample set to be confirmed.
When the sample set to be confirmed is the set, such as keyword retrieval for needing the multiple candidate samples for confirming to constitute,
Multiple candidate keywords are constituted into sample set to be confirmed, to confirm whether each candidate keywords is target keyword.
The confidence level of the sample to be confirmed is general according to application demand, using sample to be confirmed and goal verification model
Matching degree is obtained, and during such as keyword retrieval, can be built by keyword training data and be obtained keyword recognition model, is counted successively
Calculate the matching degree of keyword to be confirmed and the keyword recognition model, you can obtain the confidence level of each keyword to be confirmed;
During such as speaker verification, can be obtained with the matching degree of speaker verification's model by calculating the speech data of speaker to be confirmed
The confidence level of speaker to be confirmed.The computational methods of the confidence level can be used including the various related skill including prior art
Art, the application is not construed as limiting to the acquisition methods of the confidence level.
S12:Determine the similar sample of each sample to be confirmed in sample set to be confirmed.
S13:Confidence level according to the similar sample is modified to the confidence level of each sample to be confirmed, obtains each
The revised confidence level of sample to be confirmed.
Details are provided below.
S14:Revised confidence level according to each sample to be confirmed confirms to each sample to be confirmed.
During specific confirmation, directly judge the revised confidence level of each sample to be confirmed whether more than threshold set in advance
Value, if it is greater, then confirming successfully, i.e., sample to be confirmed is target sample;Otherwise, failure is confirmed, i.e., sample to be confirmed is not
Target sample.
If Fig. 2 is to carry out sample using confidence level after sample amendment to be confirmed to confirm schematic diagram, in sample set to be confirmed
Totally 18 samples to be confirmed, dotted line represents sample confidence threshold value line of demarcation to be confirmed, and solid rim represents positive example sample, open circle
Negative example sample is represented, the positive example sample is to be confirmed sample of the confidence level more than confidence threshold value, and the negative example sample is to put
To be confirmed sample of the reliability less than confidence threshold value;Due to being influenceed by external factor such as environment, sample is carried out using existing method
During this confirmation, easily there is the sample to be confirmed for confirming mistake, by the sample 1 and sample 2 of error check in such as Fig. 2 (a);
By herein described method, the confidence level that the similar sample according to sample to be confirmed treats confirmatory sample is repaiied
After just, revised confidence level is set to contain the information of more multisample, such as the confidence level of sample to be confirmed 1, while contain treating
Confirmatory sample 1 and its 3 information of similar sample, when carrying out sample using the revised confidence level and confirming, can be by
Confirm that the sample of mistake correctly confirms before, so as to lift the degree of accuracy of sample confirmation;
As in Fig. 2 (b), sample confirmation carried out using the revised confidence level of sample to be confirmed, can confirm by before wrong
Sample 1 and sample 2 correctly confirm by mistake.
Determination and the makeover process of confidence level below to similar sample is illustrated.
In order to improve the accuracy of sample confirmation, the application finds each sample to be confirmed from sample set to be confirmed
Similar sample;The confidence level of each sample to be confirmed is repaiied using the confidence level of the similar sample of each sample to be confirmed
Just so that revised confidence level make use of the information of the similar sample of each sample to be confirmed, the specific following institute of modification method
State.
As shown in figure 3, determine the method for the similar sample of each sample to be confirmed in sample set to be confirmed including:
S31:Calculate the similarity of each sample to be confirmed and other samples to be confirmed in sample set to be confirmed.
The similarity generally uses the description of the distance between sample to be confirmed, such as Euclidean distance, COS distance, it is described away from
From computational methods may refer to including the various correlation techniques including prior art, circular is not construed as limiting.As closed
When keyword is retrieved, by the dynamic time warping (Dynamic for calculating keyword to be confirmed and other keywords in training data
Time Warping, DTW) distance obtains;As speaker verification when, can by calculate speaker's speech data vocal print feature it
Between COS distance obtain, the vocal print feature such as Ivector features;Generally, the distance between sample is smaller, similar
Degree is bigger.
The similarity of sample to be confirmed and each sample in training data can certainly be described using other methods, it is such as straight
Connect and sample to be confirmed is matched with other samples to be confirmed, obtain matching for sample to be confirmed and other samples to be confirmed
Degree, the similarity of sample to be confirmed and other samples to be confirmed is described using the matching degree.
During specific calculating, each sample to be confirmed is used as current sample to be confirmed during sample set to be confirmed is selected successively;
Calculate current sample to be confirmed and the similarity of other each samples to be confirmed successively again, use D (X, xj) expression, wherein X tables
Show current sample to be confirmed, xjRepresent j-th sample to be confirmed in addition to current sample to be confirmed in sample set to be confirmed;
After calculating terminates, the similarity of each sample to be confirmed and other samples to be confirmed in sample set to be confirmed is obtained.
S32:According to each sample to be confirmed and the similarity of other samples to be confirmed, the phase of each sample to be confirmed is determined
Like sample.
First the above-mentioned similarity being calculated can specifically be carried out regular, each is determined further according to the similarity after regular
The similar sample of sample to be confirmed.
When specific regular, according to institute's subject to confirmation sample in sample set to be confirmed respectively between other samples to be confirmed
The maximum and minimum value of similarity carry out regular to each sample to be confirmed and the similarity of other samples to be confirmed successively, obtain
To the similarity of each sample to be confirmed and other samples to be confirmed after regular;Shown in specific regular method such as formula (1):
Wherein, S (X, xj) it is jth after current sample to be confirmed and the current sample to be confirmed of removal in sample set to be confirmed
Individual sample xjSimilarity after regular, min (D) in sample set to be confirmed institute's subject to confirmation sample it is to be confirmed with other respectively
The minimum value of Sample Similarity, max (D) in sample set to be confirmed institute's subject to confirmation sample respectively with other samples to be confirmed
The maximum of similarity.
Obtain it is regular after similarity after, can will be greater than predetermined threshold value it is regular after similarity corresponding to other
Sample to be confirmed, is defined as the similar sample of each sample to be confirmed;Or, to similarity after regular according to from big to small
Order sort, the preceding predetermined number of selected and sorted it is regular after similarity, by select it is regular after similarity corresponding to
Other samples to be confirmed, be defined as each similar sample to be confirmed.
After the similar sample for determining each sample to be confirmed, can adopt with the following method to each sample to be confirmed
Confidence level is modified.
Specifically, held each sample to be confirmed as current sample to be confirmed, and the current sample to be confirmed of correspondence
Row following steps:According to current sample to be confirmed sample similar to each it is regular after similarity and each similar sample
Confidence level, calculates the contribution rate of all similar sample of current sample to be confirmed;
Confidence level and the contribution rate to current sample to be confirmed are weighted summation, obtain current sample to be confirmed
Revised confidence level;
Wherein, the contribution rate is the contribution degree and current sample to be confirmed of all similar sample of current sample to be confirmed
To all similar samples it is regular after similarity and ratio, the contribution degree is current sample to be confirmed sample similar to each
The sum of products of the confidence level of the similarity sample similar to each after originally regular.
It is formulated as shown in formula (2):
G (X)=(1- α) c (X)+α T (X) (2)
Wherein, g (X) is the revised confidence level of current sample to be confirmed;C (X) is the amendment of current sample to be confirmed
Preceding confidence level;T (X) is the contribution rate of all similar sample of current sample to be confirmed;α is all of current sample to be confirmed
The weight of similar sample contribution rate, can be set according to application demand.
The contribution rate T (X) be according to current sample to be confirmed sample similar to each it is regular after similarity and
What the confidence calculations of each similar sample were obtained, as shown in formula (3):
Wherein,It is the contribution degree of all similar sample of current sample to be confirmed;For
Current sample to be confirmed and all similar samples it is regular after similarity and;S(X,xi) for current sample to be confirmed with its i-th
Individual similar sample it is regular after similarity;c(xi) be current sample to be confirmed i-th confidence level of similar sample;N is to work as
The similar total sample number of preceding sample to be confirmed.
Below by taking keyword retrieval as an example, flow is illustrated to be confirmed to sample.
For example:During keyword retrieval, sample set to be confirmed is all candidate keywords of keyword " A ", such as to be confirmed
Sample set is L={ a1,a2,...,am, wherein, each element is a sample (candidate keywords) to be confirmed, and m is true to treat
Total sample number is recognized, it is necessary to confirm whether each sample to be confirmed is keyword " A ", and the confidence level of each sample to be confirmed can lead to
Decoded result when crossing keyword retrieval is obtained, specific to confirm that process is as follows:
The confidence level of each sample to be confirmed in sample set to be confirmed is obtained first;
It is then determined that in sample set to be confirmed each sample to be confirmed similar sample;
Specifically can by calculating the similarity between each sample to be confirmed, and the similarity is carried out it is regular after,
The similar sample of setting respective threshold or selection fixed number is obtained, such as sample a to be confirmed1Similar sample be { a3,a6,
a7,a10};
Recycle the similar sample of each sample to be confirmed to be modified the confidence level of each sample to be confirmed, specifically repair
Timing, the confidence level of the similar sample according to sample to be confirmed, and sample to be confirmed sample similar to each it is regular after
The confidence level that similarity treats confirmatory sample is modified, and obtains each revised confidence level of sample to be confirmed;
Sample confirmation is carried out finally according to revised confidence level, to judge whether each sample to be confirmed is keyword
" A ", specifically can be by judging whether each revised confidence level of sample to be confirmed exceedes threshold value set in advance, if super
Cross, then confirm successfully, i.e., sample to be confirmed is keyword " A ";Otherwise, failure is confirmed, i.e., sample to be confirmed is not keyword
“A”;
Additionally, herein described method can be also used for speaker verification, gender confirm etc. to need to carry out sample it is true
In the application recognized, specifically it is not construed as limiting.
During specific implementation, with reference to client and service end, flow as shown in Figure 4 is provided:
S41:Client receives the sample to be confirmed of user input.
For example, in keyword retrieval, receiving the keyword of user input;Or, in speaker verification, receive user
The speech data of input.
S42:Sample to be confirmed is sent to service end by client.
Sample to be confirmed can be sent to service end by client by the network connection between service end.
S43:Service end receives the sample to be confirmed that client sends.
The present embodiment is so that service end receives the sample to be confirmed that client sends as an example, it is to be understood that service end is also
Sample to be confirmed can be got from the database of service end or by network crawl, or, it is to be confirmed that service end is obtained
Sample can come from client with a part, and another part is from the database with service end or network crawl data.
S44:Service end determines the similar sample of each sample to be confirmed in sample set to be confirmed.
S45:Service end is modified according to the similar sample to the confidence level of each sample to be confirmed, obtains each and treats
The revised confidence level of confirmatory sample.
S46:Service end confirms according to the revised confidence level of each sample to be confirmed to each sample to be confirmed.
S47:Service end obtains feedback result according to confirmation result, and feedback result is sent into client.
Wherein, service end can will confirm that result directly as feedback result, so as to by whether the successful result for confirming is sent out
Give client;Or, feedback result can be the information related to target sample, such as when user carries out keyword retrieval,
The keyword relational information that will be retrieved is sent to client;Or, feedback result is direct after can also confirming successfully for sample
The relevant information of the subsequent treatment for carrying out, such as, when user passes through voice entry, if it is confirmed that the speech data of user is target
Sample, now shows that sample confirms successfully, then service end can directly carry out subsequent treatment after sample confirms successfully, without
Sample confirmation successful information is first sent to client;Service end after directly subsequent treatment is carried out after sample confirms successfully,
Client can be sent to using the relevant information of subsequent treatment as feedback result, such as, and when above-mentioned user passes through voice entry,
Service end confirms that the speech data of user, after target sample, then to allow User logs in, obtains the individual after user's Successful login
Login page, afterwards service end by the data is activation of personal login page to client, so that client is according to receiving
Page data the operation such as render and represents corresponding personal login page.It is understood that according to application scenarios not
With can also be other situations, the application is not limited.
S48:The feedback result is presented to user by client.
The particular content of above steps may refer to the associated description in related embodiment, will not be described in detail herein.
It is understood that above-mentioned client and service end can be located in different physical equipments respectively, such as client
In the terminal device of user side, service end is located in server at end, and terminal device passes through network connection with server;Or
Person, client and service end may be located in identical physical equipment, for example, integrated client and service end in terminal device
Function, such that it is able to terminal device locally complete sample confirm.
In the present embodiment, by determining the similar sample of sample to be confirmed, and confirmatory sample is treated according to similar sample
Confidence level is modified, and carries out sample confirmation according to revised confidence level, can effectively lift the accurate of sample confirmation
Degree, especially corrects sample of the previous belief near confidence threshold value, and lifting effect becomes apparent from.
Fig. 5 is the structural representation that the sample that the application one embodiment is proposed confirms device.
As shown in figure 5, the device 50 of the present embodiment includes:Acquisition module 51, determining module 52, correcting module 53 and confirmation
Module 54.
Acquisition module 51, the confidence level for obtaining each sample to be confirmed in sample set to be confirmed;
Determining module 52, the similar sample for determining each sample to be confirmed in sample set to be confirmed;
Correcting module 53, for being modified to the confidence level of each sample to be confirmed according to the similar sample, obtains
The revised confidence level of each sample to be confirmed;
Confirm module 54, each sample to be confirmed is carried out for the revised confidence level according to each sample to be confirmed
Confirm.
In some embodiments, referring to Fig. 6, the determining module 52 includes:
Calculating sub module 521, for calculating each sample to be confirmed and other samples to be confirmed in sample set to be confirmed
Similarity;
Determination sub-module 522, for the similarity according to each sample to be confirmed and other samples to be confirmed, determines each
The similar sample of sample to be confirmed.
In some embodiments, the determination sub-module 522 specifically for:
Each sample to be confirmed and the similarity of other samples to be confirmed are carried out it is regular, obtain it is regular after similarity;
According to the similarity after regular, the similar sample of each sample to be confirmed is determined.
In some embodiments, the determination sub-module 522 is used to, according to the similarity after regular, determine each sample to be confirmed
This similar sample, including:
Will be greater than predetermined threshold value it is regular after similarity corresponding to other samples to be confirmed, be defined as each to be confirmed
The similar sample of sample;Or,
Similarity after regular is sorted according to order from big to small, the preceding predetermined number of selected and sorted it is regular after
Similarity, by select it is regular after similarity corresponding to other samples to be confirmed, be defined as each to be confirmed similar
Sample.
In some embodiments, the correcting module 53 specifically for:
Using each sample to be confirmed as current sample to be confirmed, and the current sample to be confirmed of correspondence performs following step
Suddenly:
According to current sample to be confirmed sample similar to each it is regular after similarity and each similar sample put
Reliability, calculates the contribution rate of all similar sample of current sample to be confirmed;
Confidence level and the contribution rate to current sample to be confirmed are weighted summation, obtain current sample to be confirmed
Revised confidence level;
Wherein, the contribution rate is the contribution degree and current sample to be confirmed of all similar sample of current sample to be confirmed
To all similar samples it is regular after similarity and ratio, the contribution degree is current sample to be confirmed sample similar to each
The sum of products of the confidence level of the similarity sample similar to each after originally regular.
It is understood that the device of the present embodiment is corresponding with above method embodiment, particular content may refer to method
The associated description of embodiment, no longer describes in detail herein.
In the present embodiment, by determining the similar sample of sample to be confirmed, and confirmatory sample is treated according to similar sample
Confidence level is modified, and carries out sample confirmation according to revised confidence level, can effectively lift the accurate of sample confirmation
Degree, especially corrects sample of the previous belief near confidence threshold value, and lifting effect becomes apparent from.
Fig. 7 is the structural representation that the sample that the application one embodiment is proposed confirms system.
As shown in fig. 7, the system of the present embodiment includes:Client 71 and service end 72.
Client 71, the sample to be confirmed for receiving user input;
Service end 72, the sample to be confirmed for receiving client transmission, each is treated really in obtaining sample set to be confirmed
Recognize the confidence level of sample;Determine the similar sample of each sample to be confirmed in sample set to be confirmed;According to the similar sample
Confidence level to each sample to be confirmed is modified, and obtains the revised confidence level of each sample to be confirmed;According to each
The revised confidence level of sample to be confirmed confirms to each sample to be confirmed.
In some embodiments, the service end 72 is additionally operable to:According to confirming that result obtains feedback result, and by the feedback
Result is sent to client;
The client 71 is additionally operable to:The feedback result that the service end sends is received, and the feedback result is fed back
To user.
In Fig. 7 so that client is connected with service end by wireless network as an example, it is to be understood that client and service end
Can also be connected by cable network, or, if client and service end are integrated in same equipment, client and service end
Can be connected by the bus of device interior.
It is understood that the function of service end is consistent with above-mentioned device, therefore, the concrete composition of service end can join
See the device shown in Fig. 5 or Fig. 6, will not be described in detail herein.
In the present embodiment, by determining the similar sample of sample to be confirmed, and confirmatory sample is treated according to similar sample
Confidence level is modified, and carries out sample confirmation according to revised confidence level, can effectively lift the accurate of sample confirmation
Degree, especially corrects sample of the previous belief near confidence threshold value, and lifting effect becomes apparent from.
It is understood that same or similar part can mutually refer in the various embodiments described above, in certain embodiments
Unspecified content may refer to same or analogous content in other embodiment.
It should be noted that in the description of the present application, term " first ", " second " etc. are only used for describing purpose, without
It is understood that to indicate or implying relative importance.Additionally, in the description of the present application, unless otherwise indicated, the implication of " multiple "
Refer at least two.
Any process described otherwise above or method description in flow chart or herein is construed as, and expression includes
It is one or more for realizing specific logical function or process the step of the module of code of executable instruction, fragment or portion
Point, and the scope of the preferred embodiment of the application includes other realization, wherein can not press shown or discussion suitable
Sequence, including function involved by basis by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be by the application
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each several part of the application can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In implementation method, the software that multiple steps or method can in memory and by suitable instruction execution system be performed with storage
Or firmware is realized.If for example, realized with hardware, and in another embodiment, can be with well known in the art
Any one of row technology or their combination are realized:With the logic gates for realizing logic function to data-signal
Discrete logic, the application specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method is carried
The rapid hardware that can be by program to instruct correlation is completed, and described program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
Additionally, during each functional unit in the application each embodiment can be integrated in a processing module, it is also possible to
It is that unit is individually physically present, it is also possible to which two or more units are integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, it would however also be possible to employ the form of software function module is realized.The integrated module is such as
Fruit is to realize in the form of software function module and as independent production marketing or when using, it is also possible to which storage is in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means to combine specific features, structure, material or spy that the embodiment or example are described
Point is contained at least one embodiment of the application or example.In this manual, to the schematic representation of above-mentioned term not
Necessarily refer to identical embodiment or example.And, the specific features of description, structure, material or feature can be any
One or more embodiments or example in combine in an appropriate manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is impossible to the limitation to the application is interpreted as, one of ordinary skill in the art within the scope of application can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (12)
1. a kind of sample confirmation method, it is characterised in that including:
Obtain the confidence level of each sample to be confirmed in sample set to be confirmed;
Determine the similar sample of each sample to be confirmed in sample set to be confirmed;
The confidence level of each sample to be confirmed is modified according to the similar sample, obtains the amendment of each sample to be confirmed
Confidence level afterwards;
Revised confidence level according to each sample to be confirmed confirms to each sample to be confirmed.
2. method according to claim 1, it is characterised in that each sample to be confirmed in the determination sample set to be confirmed
This similar sample, including:
Calculate the similarity of each sample to be confirmed and other samples to be confirmed in sample set to be confirmed;
According to each sample to be confirmed and the similarity of other samples to be confirmed, the similar sample of each sample to be confirmed is determined.
3. method according to claim 2, it is characterised in that described according to each sample to be confirmed and other samples to be confirmed
This similarity, determines the similar sample of each sample to be confirmed, including:
Each sample to be confirmed and the similarity of other samples to be confirmed are carried out it is regular, obtain it is regular after similarity;
According to the similarity after regular, the similar sample of each sample to be confirmed is determined.
4. method according to claim 3, it is characterised in that the similarity according to after regular, determines that each is treated really
Recognize the similar sample of sample, including:
Will be greater than predetermined threshold value it is regular after similarity corresponding to other samples to be confirmed, be defined as each sample to be confirmed
Similar sample;Or,
Similarity after regular is sorted according to order from big to small, the preceding predetermined number of selected and sorted it is regular after phase
Like spend, by select it is regular after similarity corresponding to other samples to be confirmed, be defined as each similar sample to be confirmed.
5. method according to claim 1, it is characterised in that it is described according to the similar sample to each sample to be confirmed
Confidence level be modified, obtain the revised confidence level of each sample to be confirmed, including:
Using each sample to be confirmed as current sample to be confirmed, and the current sample to be confirmed of correspondence performs following steps:
According to current sample to be confirmed sample similar to each it is regular after similarity and each similar sample confidence level,
Calculate the contribution rate of all similar sample of current sample to be confirmed;
Confidence level and the contribution rate to current sample to be confirmed are weighted summation, obtain the amendment of current sample to be confirmed
Confidence level afterwards;
Wherein, the contribution rate is contribution degree and current sample to be confirmed and the institute of all similar sample of current sample to be confirmed
Have similar sample it is regular after similarity sum ratio, the contribution degree is current sample to be confirmed sample similar to each
The sum of products of the confidence level of the similarity sample similar to each after regular.
6. a kind of sample confirms device, it is characterised in that including:
Acquisition module, the confidence level for obtaining each sample to be confirmed in sample set to be confirmed;
Determining module, the similar sample for determining each sample to be confirmed in sample set to be confirmed;
Correcting module, for being modified to the confidence level of each sample to be confirmed according to the similar sample, obtains each and treats
The revised confidence level of confirmatory sample;
Confirm module, each sample to be confirmed is confirmed for the revised confidence level according to each sample to be confirmed.
7. device according to claim 6, it is characterised in that the determining module includes:
Calculating sub module, for calculating sample set to be confirmed in each sample to be confirmed it is similar to other samples to be confirmed
Degree;
Determination sub-module, for the similarity according to each sample to be confirmed and other samples to be confirmed, determines that each is to be confirmed
The similar sample of sample.
8. device according to claim 7, it is characterised in that the determination sub-module specifically for:
Each sample to be confirmed and the similarity of other samples to be confirmed are carried out it is regular, obtain it is regular after similarity;
According to the similarity after regular, the similar sample of each sample to be confirmed is determined.
9. device according to claim 8, it is characterised in that the determination sub-module is used for according to similar after regular
Degree, determines the similar sample of each sample to be confirmed, including:
Will be greater than predetermined threshold value it is regular after similarity corresponding to other samples to be confirmed, be defined as each sample to be confirmed
Similar sample;Or,
Similarity after regular is sorted according to order from big to small, the preceding predetermined number of selected and sorted it is regular after phase
Like spend, by select it is regular after similarity corresponding to other samples to be confirmed, be defined as each similar sample to be confirmed.
10. device according to claim 6, it is characterised in that the correcting module specifically for:
Using each sample to be confirmed as current sample to be confirmed, and the current sample to be confirmed of correspondence performs following steps:
According to current sample to be confirmed sample similar to each it is regular after similarity and each similar sample confidence level,
Calculate the contribution rate of all similar sample of current sample to be confirmed;
Confidence level and the contribution rate to current sample to be confirmed are weighted summation, obtain the amendment of current sample to be confirmed
Confidence level afterwards;
Wherein, the contribution rate is contribution degree and current sample to be confirmed and the institute of all similar sample of current sample to be confirmed
Have similar sample it is regular after similarity sum ratio, the contribution degree is current sample to be confirmed sample similar to each
The sum of products of the confidence level of the similarity sample similar to each after regular.
A kind of 11. samples confirm system, it is characterised in that including:
Client, the sample to be confirmed for receiving user input;
Service end, the sample to be confirmed for receiving client transmission, obtains each sample to be confirmed in sample set to be confirmed
Confidence level;Determine the similar sample of each sample to be confirmed in sample set to be confirmed;According to the similar sample to each
The confidence level of sample to be confirmed is modified, and obtains the revised confidence level of each sample to be confirmed;It is to be confirmed according to each
The revised confidence level of sample confirms to each sample to be confirmed.
12. systems according to claim 11, it is characterised in that
The service end is additionally operable to:Feedback result is obtained according to confirmation result, and the feedback result is sent to client;
The client is additionally operable to:The feedback result that the service end sends is received, and the feedback result is fed back into user.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710031626.2A CN106708806B (en) | 2017-01-17 | 2017-01-17 | Sample confirmation method, device and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710031626.2A CN106708806B (en) | 2017-01-17 | 2017-01-17 | Sample confirmation method, device and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106708806A true CN106708806A (en) | 2017-05-24 |
CN106708806B CN106708806B (en) | 2020-06-02 |
Family
ID=58908535
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710031626.2A Active CN106708806B (en) | 2017-01-17 | 2017-01-17 | Sample confirmation method, device and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106708806B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111199728A (en) * | 2018-10-31 | 2020-05-26 | 阿里巴巴集团控股有限公司 | Training data acquisition method and device, intelligent sound box and intelligent television |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
CN103530604A (en) * | 2013-09-27 | 2014-01-22 | 中国人民解放军空军工程大学 | Robustness visual tracking method based on transductive effect |
CN103984738A (en) * | 2014-05-22 | 2014-08-13 | 中国科学院自动化研究所 | Role labelling method based on search matching |
CN104268900A (en) * | 2014-09-26 | 2015-01-07 | 中安消技术有限公司 | Motion object detection method and device |
CN104392439A (en) * | 2014-11-13 | 2015-03-04 | 北京智谷睿拓技术服务有限公司 | Image similarity confirmation method and device |
-
2017
- 2017-01-17 CN CN201710031626.2A patent/CN106708806B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102760285A (en) * | 2012-05-31 | 2012-10-31 | 河海大学 | Image restoration method |
CN103530604A (en) * | 2013-09-27 | 2014-01-22 | 中国人民解放军空军工程大学 | Robustness visual tracking method based on transductive effect |
CN103984738A (en) * | 2014-05-22 | 2014-08-13 | 中国科学院自动化研究所 | Role labelling method based on search matching |
CN104268900A (en) * | 2014-09-26 | 2015-01-07 | 中安消技术有限公司 | Motion object detection method and device |
CN104392439A (en) * | 2014-11-13 | 2015-03-04 | 北京智谷睿拓技术服务有限公司 | Image similarity confirmation method and device |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111199728A (en) * | 2018-10-31 | 2020-05-26 | 阿里巴巴集团控股有限公司 | Training data acquisition method and device, intelligent sound box and intelligent television |
Also Published As
Publication number | Publication date |
---|---|
CN106708806B (en) | 2020-06-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9542948B2 (en) | Text-dependent speaker identification | |
US20150127336A1 (en) | Speaker verification using neural networks | |
US20130054236A1 (en) | Method for the detection of speech segments | |
CN106649694A (en) | Method and device for identifying user's intention in voice interaction | |
CN110827833A (en) | Segment-based speaker verification using dynamically generated phrases | |
CN1302427A (en) | Model adaptation system and method for speaker verification | |
US20170263241A1 (en) | Apparatus and method for training a neutral network acoustic model, and speech recognition apparatus and method | |
CN107291775B (en) | Method and device for generating repairing linguistic data of error sample | |
CN109243466A (en) | A kind of vocal print authentication training method and system | |
CN111382270A (en) | Intention recognition method, device and equipment based on text classifier and storage medium | |
US10916254B2 (en) | Systems, apparatuses, and methods for speaker verification using artificial neural networks | |
CN108509727B (en) | Model selection processing method and device in data modeling | |
Zhang et al. | I-vector Transformation Using Conditional Generative Adversarial | |
CN113870860A (en) | End-to-end voiceprint recognition method and voiceprint recognition device | |
Sari et al. | Auxiliary networks for joint speaker adaptation and speaker change detection | |
CN115312033A (en) | Speech emotion recognition method, device, equipment and medium based on artificial intelligence | |
CN104901807A (en) | Vocal print password method available for low-end chip | |
CN106708806A (en) | Sample determination method, device and system | |
US7509257B2 (en) | Method and apparatus for adapting reference templates | |
CN111402899B (en) | Cross-channel voiceprint recognition method and device | |
CN116257816A (en) | Accompanying robot emotion recognition method, device, storage medium and equipment | |
CN112131415B (en) | Method and device for improving data acquisition quality based on deep learning | |
WO2021257000A1 (en) | Cross-modal speaker verification | |
US7085718B2 (en) | Method for speaker-identification using application speech | |
CN113656566A (en) | Intelligent dialogue processing method and device, computer equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |