CN110389666A - A kind of input error correction method and device - Google Patents
A kind of input error correction method and device Download PDFInfo
- Publication number
- CN110389666A CN110389666A CN201810339438.0A CN201810339438A CN110389666A CN 110389666 A CN110389666 A CN 110389666A CN 201810339438 A CN201810339438 A CN 201810339438A CN 110389666 A CN110389666 A CN 110389666A
- Authority
- CN
- China
- Prior art keywords
- input
- error correction
- coded strings
- position sequence
- input position
- 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
- 238000012937 correction Methods 0.000 title claims abstract description 340
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000015654 memory Effects 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 14
- 235000013399 edible fruits Nutrition 0.000 claims description 5
- 238000013459 approach Methods 0.000 abstract description 8
- 238000010586 diagram Methods 0.000 description 11
- 238000004891 communication Methods 0.000 description 10
- 230000006399 behavior Effects 0.000 description 6
- 230000009471 action Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 108091092195 Intron Proteins 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000013136 deep learning model Methods 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000712 assembly Effects 0.000 description 2
- 238000000429 assembly Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- KLDZYURQCUYZBL-UHFFFAOYSA-N 2-[3-[(2-hydroxyphenyl)methylideneamino]propyliminomethyl]phenol Chemical compound OC1=CC=CC=C1C=NCCCN=CC1=CC=CC=C1O KLDZYURQCUYZBL-UHFFFAOYSA-N 0.000 description 1
- 241000721047 Danaus plexippus Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000010009 beating Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000001351 cycling effect Effects 0.000 description 1
- 201000001098 delayed sleep phase syndrome Diseases 0.000 description 1
- 208000033921 delayed sleep phase type circadian rhythm sleep disease Diseases 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000012092 media component Substances 0.000 description 1
- 235000012149 noodles Nutrition 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/02—Input arrangements using manually operated switches, e.g. using keyboards or dials
- G06F3/023—Arrangements for converting discrete items of information into a coded form, e.g. arrangements for interpreting keyboard generated codes as alphanumeric codes, operand codes or instruction codes
- G06F3/0233—Character input methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/232—Orthographic correction, e.g. spell checking or vowelisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Input From Keyboards Or The Like (AREA)
Abstract
The embodiment of the present application discloses a kind of input error correction method, which comprises obtains the input position sequence of coded strings and the input coded strings;By the input position sequence inputting error correcting model, the error correcting model is obtained according to the training of history input position sequence;If judging there is wrong input coding in the coded strings, the error correcting model output is directed to the error correction result of the coded strings;Error correction is carried out for the coded strings according to the error correction result.It can be seen that, since the error-correction rule in error correcting model is to be trained according to history input position sequence, learn to obtain, it avoids and artificially presets error-correction rule and the stiff problem of general rule in traditional approach, error correction efficiently can be carried out to coded strings, when for complicated error correction situation, it does not need as called a large amount of error-correction rule in traditional approach or in order to save make-up time discard portion error-correction rule, not only reduces the make-up time, and can be improved error correction accuracy and comprehensive.
Description
Technical field
This application involves input method fields, more particularly to a kind of input error correction method and device.
Background technique
User often will appear when using input method and input spelling under the case where inputing coded strings by mistake, such as spelling input method by mistake
Sound string.Coded strings can there are many other wrong defeated codings caused by the possibility inputed by mistake, such as the smaller click offset of input keyboard, such as
Rapidly input that lead to encode tandem reverse etc..Input method can carry out corresponding error correction for the behavior of inputing by mistake of user, entangle
Wrong foundation is usually the rule being manually set, and various error-correction rules may be arranged for different inputing by mistake.
However, when being directed to more complicated error correction situation, such as there is a variety of input by mistake in the coded strings that are inputted of user
May or it is multiple when inputing coding by mistake, since the error-correction rule being artificially arranged is general more and many and diverse, cause to input by mistake for a variety of
May or it is multiple when inputing code correction by mistake, need that a large amount of error-correction rule is called to carry out error correction.Error correction is carried out in this way or is caused
Make-up time is too long or causes error correction result inaccurate or incomplete to save make-up time discard portion error-correction rule.
User is made when for more complicated error correction situation as it can be seen that carrying out error correction by the error-correction rule being artificially arranged
It is experienced at bad error correction.
Summary of the invention
It in order to solve the above-mentioned technical problem, being capable of Active Learning this application provides a kind of input error correction method and device
Rule eliminates the influence of artificial setting rule, improves error correction accuracy and comprehensive.
The embodiment of the present application discloses following technical solution:
In a first aspect, the embodiment of the present application provides a kind of input error correction method, which comprises
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is according to the training of history input position sequence
It obtains;
If judging that there is wrong input coding in the coded strings, error correcting model output the entangling for the coded strings
Wrong result;
Error correction is carried out for the coded strings according to the error correction result.
Optionally, the history input position sequence includes that the corresponding input position of coded strings is correctly entered in historical data
Sequence input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched volume of character in history with corresponding in history input coding string
Sequence.
Optionally, the input position sequence includes input bit confidence when each coding in the coded strings is entered
Breath;The input position information includes screen area information or key information.
Optionally, the screen area information includes screen area location information or screen area pictorial information;It is described
The region that screen area is clicked when clicking coding on the screen by user;Wherein, the screen area location information includes institute
State the center position coordinates of all pixels coordinate or the screen area in screen area.
Optionally, the key information includes the key that user is clicked when input coding in physical keyboard.
Optionally, the error correcting model is to be obtained according to the combined training of any one or more in following manner:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;It is incited somebody to action described
Before the input position sequence inputting error correcting model, further includes:
Obtain the corresponding target input scene of the coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target input scene described in the input position sequence inputting;
The second way are as follows:
The history input position sequence that is inputted according to the error correcting model by same type of input keyboard is trained
It arrives;It is described by the input position sequence inputting error correcting model before, further includes:
Obtain the corresponding target keyboard type of the input position sequence;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target keyboard type described in the input position sequence inputting;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;It is described to obtain
Take the input position sequence of coded strings and the input coded strings, comprising:
Obtain the input position sequence and time interval sequence of the coded strings and the input coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the input position sequence and the time interval sequence inputting error correcting model.
Optionally, the error correction result that the coded strings are directed to by error correcting model output, comprising:
It is corresponding credible for the error correction result of the coded strings and the error correction result by error correcting model output
Spend score;Wherein, the confidence level score is used to indicate the credibility of the error correction result;
It is described that error correction is carried out for the coded strings according to the error correction result, comprising:
If the confidence level score of the error correction result is greater than threshold value, according to the error correction result and the error correction result
Confidence level score carries out error correction for the coded strings.
Second aspect, the embodiment of the present application provide a kind of input error correction device, and described device includes: acquiring unit, defeated
Enter unit, output unit and error correction unit;
The acquiring unit, for obtaining the input position sequence of coded strings and the input coded strings;
The input unit, for being gone through according to the error correcting model by the input position sequence inputting error correcting model
The training of history input position sequence obtains;
The output unit, if for judging that there is wrong input coding in the coded strings, the error correcting model output
For the error correction result of the coded strings;
The error correction unit, for carrying out error correction for the coded strings according to the error correction result.
Optionally, the history input position sequence includes that the corresponding input position of coded strings is correctly entered in historical data
Sequence input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched volume of character in history with corresponding in history input coding string
Sequence.
Optionally, the input position sequence includes input bit confidence when each coding in the coded strings is entered
Breath;The input position information includes screen area information or key information.
Optionally, the screen area information includes screen area location information or screen area pictorial information;It is described
The region that screen area is clicked when clicking coding on the screen by user;Wherein, the screen area location information includes institute
State the center position coordinates of all pixels coordinate or the screen area in screen area.
Optionally, the key information includes the key that user is clicked when input coding in physical keyboard.
Optionally, the error correcting model is to be obtained according to the combined training of any one or more in following manner:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;
The acquiring unit is also used to obtain the corresponding target input scene of the coded strings;
The input unit is also used to the corresponding error correction mould of target input scene described in the input position sequence inputting
Type;
The second way are as follows:
The history input position sequence that is inputted according to the error correcting model by same type of input keyboard is trained
It arrives;
The acquiring unit is also used to obtain the corresponding target keyboard type of the input position sequence;
The input unit is also used to the corresponding error correction mould of target keyboard type described in the input position sequence inputting
Type;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;
The acquiring unit, be also used to obtain the coded strings and the input coded strings input position sequence and when
Between intervening sequence;
The input unit is also used to the input position sequence and the time interval sequence inputting error correcting model.
Optionally, the output unit is also used to export the error correction knot for being directed to the coded strings by the error correcting model
Fruit and the corresponding confidence level score of the error correction result;Wherein, what the confidence level score was used to indicate the error correction result can
Letter degree;
The error correction unit, if the confidence level score for being also used to the error correction result is greater than threshold value, according to the error correction
As a result with the confidence level score of the error correction result, error correction is carried out for the coded strings.
The third aspect, the embodiment of the present application provide a kind of input error correction apparatus, and the equipment includes memory, and
One perhaps more than one program one of them or more than one program be stored in memory, and be configured to by one
It includes the instruction for performing the following operation that a or more than one processor, which executes the one or more programs:
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is according to the training of history input position sequence
It obtains;
If judging that there is wrong input coding in the coded strings, error correcting model output the entangling for the coded strings
Wrong result;
Error correction is carried out for the coded strings according to the error correction result.
Fourth aspect, the embodiment of the present application provide a kind of non-transitorycomputer readable storage medium, when the storage
When instruction in medium is executed by the processor of electronic equipment, so that electronic equipment is able to carry out a kind of input error correction method, institute
The method of stating includes:
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is according to the training of history input position sequence
It obtains;
If judging that there is wrong input coding in the coded strings, error correcting model output the entangling for the coded strings
Wrong result;
Error correction is carried out for the coded strings according to the error correction result.
It can be seen from above-mentioned technical proposal when detecting coding input behavior, obtains coded strings and input the volume
The input position sequence of sequence, and by the input position sequence inputting into preset error correcting model.Since the error correcting model is
It is trained according to history input position sequence, therefore is eliminated by the error-correction rule that the error correcting model obtains and artificially set
The influence of rule is set, moreover, the training error correcting model is history input position sequence, which is user
The position sequence pressed on the input when input coding string, personalized input when can more embody user's input are special
Sign.If the error correcting model judges there is wrong input coding in the coded strings, it can export to be directed to by the error correcting model and be somebody's turn to do
Coded strings error correction result, so as to carry out error correction for the coded strings according to the error correction result.As it can be seen that due to error correcting model energy
Enough more personalized input feature vectors embodied when user inputs, so as to realize the personalized error correcting for being directed to user, moreover,
Since the error-correction rule in error correcting model is to be trained according to history input position sequence, learn to obtain, tradition is avoided
Error-correction rule and the stiff problem of general rule are artificially preset in mode, error correction efficiently can be carried out to coded strings, for
When complicated error correction situation, do not need as called a large amount of error-correction rule in traditional approach or in order to save make-up time discard portion
Error-correction rule not only reduces the make-up time, and can be improved error correction accuracy and comprehensive.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of application without any creative labor, may be used also for those of ordinary skill in the art
To obtain other drawings based on these drawings.
Fig. 1 is the block schematic illustration of an exemplary application scene provided by the embodiments of the present application;
Fig. 2 is the schematic diagram of an exemplary application scene provided by the embodiments of the present application;
Fig. 3 is a kind of method flow diagram for inputting error correction method provided by the embodiments of the present application;
Fig. 4 is the schematic diagram of an exemplary application scene provided by the embodiments of the present application;
Fig. 5 is that a kind of interface of input method provided by the embodiments of the present application shows schematic diagram;
Fig. 6 is that a kind of interface of input method provided by the embodiments of the present application shows schematic diagram;
Fig. 7 is a kind of structural schematic diagram for inputting error correction device provided by the embodiments of the present application;
Fig. 8 is a kind of structural schematic diagram for inputting error correction apparatus provided by the embodiments of the present application.
Fig. 9 is a kind of structural schematic diagram of server provided by the embodiments of the present application.
Specific embodiment
With reference to the accompanying drawing, embodiments herein is described.
User often will appear when using input method and input spelling under the case where inputing coded strings by mistake, such as spelling input method by mistake
Sound string.Coded strings can there are many other wrong defeated codings caused by the possibility inputed by mistake, such as the smaller click offset of input keyboard, such as
It is reverse etc. to rapidly input lower coding tandem.
Inventors have found that traditional input method can carry out corresponding error correction for the behavior of inputing by mistake of user, and error correction
According to being usually the rule being manually set, various error-correction rules may can be set for different inputing by mistake.However, when for compared with
There is a variety of input by mistake may or multiple input coding by mistake when for complicated error correction situation, such as in the coded strings that are inputted of user
When, since the error-correction rule that is artificially arranged is general more and many and diverse, cause for it is a variety of input by mistake may or multiple codings of inputing by mistake entangle
It staggers the time, needs that a large amount of error-correction rule is called to carry out error correction.But a large amount of error-correction rule is called to carry out error correction, it will lead to and entangle
Wrong overlong time.If being needed in error correction procedure to save the make-up time, calling section error-correction rule, and give up another
A part of error-correction rule;However, such that for error correction error-correction rule it is not comprehensive enough, cause error correction result inaccuracy or
It is incomplete.User is caused when for more complicated error correction situation as it can be seen that carrying out error correction by the error-correction rule being artificially arranged
Bad error correction experience.
For this purpose, the embodiment of the present application provides a kind of solution regarding to the issue above, in the method, when detecting
When coding input behavior, the coded strings of available input and the input position sequence for inputting the coded strings, and can should
In input position sequence inputting error correcting model.It is according to the training of history input position sequence by error correcting model in this present embodiment
It obtains, so the error correcting model can learn more comprehensive error-correction rule, so as to remove the shadow of artificial setting rule
It rings.Moreover, the training error correcting model is history input position sequence, since the history input position sequence being capable of more bodies
Therefore input feature vector when current family input coding string according to the error correcting model, can more efficiently pass through input position sequence
Column judge there is wrong input coding in the coded strings, and the error correction for being directed to the coded strings can be exported by the error correcting model
As a result, so as to carry out error correction for the coded strings according to the error correction result.In this way, can not only be efficiently to the volume of input
Sequence carries out error correction, and when for complicated error correction situation, it may not be necessary to as called a large amount of error-correction rule in traditional approach
Or in order to save make-up time discard portion error-correction rule, the influence of artificial setting rule is eliminated, so that not only can be with
Reduce the make-up time, and can be improved error correction accuracy and comprehensive.
It should be noted that the error correcting model in the embodiment of the present application can be set in the terminal device for being equipped with input method
In, such as the equipment such as mobile phone, ipad, desktop computer, it is of course also possible to be arranged in the server.Next, will be with error correcting model
It is illustrated for setting in the server.
As an example, a kind of input error correction method provided by the embodiment of the present application can be applied to as shown in Figure 1
Scene.In this scenario, error correcting model is arranged in server 103.Firstly, user 101 can use in terminal device 102
Input method input coding string.Then, terminal device 102 can respond user 101 input operation, get the coded strings with
And user 101 inputs input position sequence corresponding when the coded strings in terminal device 102, and by the input position sequence
It is sent to server 103.It then, can be by the input position sequence inputting after server 103 gets the input position sequence
Into error correcting model, if judging that there is wrong input coding, the error correcting model in the coded strings according to the input position sequence
The error correction result for the coded strings can be exported.And then, server 103 can be by the error correction result to terminal device 102
It sends.Finally, terminal device 102 can carry out error correction for the coded strings according to the error correction result, the coding after obtaining error correction
String perhaps candidate item and can to user 101 show the error correction after coded strings or candidate item.
It is understood that in above-mentioned application scenarios, although being by servicing by the action description of the application embodiment
Device 103 executes, but these movements can also part executed, partially executed by server 103 or complete by terminal device 102
It is executed entirely by terminal device 102.The application is unrestricted in terms of executing subject, as long as it is public to perform the application embodiment institute
The movement opened.
It should be noted that above-mentioned application scenarios are merely for convenience of understanding the application and showing, the embodiment party of the application
Formula is unrestricted in this regard.On the contrary, presently filed embodiment can be applied to applicable any scene.
For convenient for technical solution provided herein easier to understand, next, will be to institute in the embodiment of the present application
The some concepts being related to are explained.
Error correcting model is introduced in the embodiment of the present application, which can be according to the training of history input position sequence
It obtains, which can be the corresponding input position sequence of input coding string in historical data.In this way, can
By the input position sequence to the error correcting model input coding string, to obtain the error correction result for the coded strings.For example, false
If user wants to input Chinese character " today " in mobile phone, coded strings " juntian " is had input in mobile phone, at this point it is possible to obtain
Get coded strings " juntian " and the corresponding input position sequence of coded strings " juntian ", and by the coded strings
In " juntian " corresponding input position sequence inputting error correcting model;The error correction that the error correcting model can learn in advance according to it
Rule judges have mistake defeated in the coded strings " juntian " by the corresponding input position sequence of coded strings " juntian "
Enter coding, and the error correction result for the coded strings " juntian " can be exported, as the coded strings " juntian " are corresponding
Correct coding string " jintian " or the corresponding candidate item of correct coding string " jintian ".
Wherein, coded strings may include multiple codings.For example, in the case where input method is Chinese character coding input method, input
Coding can be phonetic alphabet, and correspondingly, coded strings can be pinyin string;And in the case where input method is English input method,
The coding of input can be English alphabet, and correspondingly, coded strings can be English alphabet string.For ease of description, in reality later
Apply in example, by mainly using input be encoded to phonetic alphabet and coded strings as phonetic alphabet string for be illustrated.
And to can be each coding in the coded strings corresponding when being entered for the input position sequence of input coding string
Input position information carries out arranging obtained input position combination according to the input sequence of each coding, wherein a coding
Corresponding input position information, which can be, is able to reflect user to input this coding institute's click location in input area
Position feature, for example can be user for the coordinate information of input coding and input area institute contact position.That is, defeated
Entering position sequence may include input position information when each coding in coded strings is entered.Next, for example bright
" input position sequence ", it is assumed that input coding string is " lima ", then corresponding input coordinate when can be entered coding " l "
Information A, coding " i " input coordinate information B corresponding when being entered, coding " m " input coordinate letter corresponding when being entered
Breath C and coding " a " input coordinate information D corresponding when being entered (encode " l " → volume according to the input sequence of each coding
Code " i " → coding " m " → coding " a ") it is ranked up, obtain input coordinate information combination " input coordinate information A → input coordinate
Information B → input coordinate information C → input coordinate information D ", and the input coordinate information can be combined and be used as coded strings
" lima " corresponding input position sequence.
It should be noted that it may include correct in historical data for being used to train the history input position sequence of error correcting model
The corresponding input position sequence of input coding string and/or the corresponding input position sequence of wrong input coding string, separately below into
Row is specific to be introduced:
In one implementation, the history input position sequence for training error correcting model may include in historical data
Correctly enter the corresponding input position sequence of coded strings.Wherein, coded strings are correctly entered and can be and are matched with upper screen character symbol
History input coding string.Continue by taking Chinese character " today " as an example, it is assumed that the upper screen character symbol that user inputs in mobile phone is that Chinese character is " modern
It ", and the coded strings inputted in mobile phone are " jintian " (i.e. history input coding string), since the upper screen character accords with " today "
Match with the coded strings " jintian ", then the coded strings " jintian " can be used as and correctly enter coded strings.
In one implementation, the history input position sequence for training error correcting model may include in historical data
The corresponding input position sequence of mistake input coding string.Wherein, mistake input coding string be history input coding string in it is corresponding
Shield the unmatched coded strings of character in history, continues by taking Chinese character " today " as an example, it is assumed that the upper screen character that user inputs in mobile phone
Symbol is Chinese character " today ", and the coded strings inputted in mobile phone are " jontian " (i.e. history input coding string), due to the coding
The character " the present " in coded strings " jon " and upper screen character symbol " today " in string " jontian " mismatches, then can be by the coding
" jon " is gone here and there as wrong input coding string.
In one implementation, the history input position sequence for training error correcting model may include correctly entering volume
The corresponding input position sequence of sequence input position sequence corresponding with wrong input coding string.Using correctly entering coded strings pair
The input position sequence answered input position sequence corresponding with wrong input coding string is trained error correcting model, can make
The error-correction rule that error correcting model after training learns is more comprehensive, to improve the error correction accuracy of error correcting model and complete
Face property.In this way, when in the coded strings that user is inputted have it is multiple input coding by mistake when, which can export for the volume
Multiple error correction results for inputing coding by mistake in sequence.
In addition, in the present embodiment, the keyboard for input coding string can be dummy keyboard, it is also possible to entity key
Disk.Also, the input mode of both keyboards is different, for example, the mode of input coding is when keyboard is dummy keyboard
User clicks a certain region in screen, and when keyboard is physical keyboard, the mode of input coding is in user's point beating keyboard
A certain key.Correspondingly, the input bit confidence in the input mode of different types of keyboard, when the coding got is entered
Breath also can be different, i.e., when the type difference of the keyboard for input coding string, input the input position sequence of error correcting model
It can be different.
Specifically, the input position when keyboard is physical keyboard, in input position sequence, when each coding is entered
Information may include key information, wherein key information can be understood as user and be clicked when input coding in physical keyboard
Key;For example, user wishes input coding " n ", and user clicks key " N " using finger in physical keyboard, then,
In input position sequence, the key information (i.e. input position information) when coding " n " is entered can be key " N ".
When keyboard is dummy keyboard, in input position sequence, input position information when each coding is entered can
To include screen area information;Wherein, screen area information can be understood as location information of the screen area in screen.Due to
When user clicks coding on the screen, the limbs of user can be contacted with multiple pixels in screen, so, it can will be with user's limb
The all pixels area defined that body is in contact is referred to as screen area, that is to say, that screen area can exist for user
The region clicked when clicking coding on screen.
It wherein, will be mainly to be used for the keyboard of input coding as dummy keyboard for ease of description, in embodiment later
For be illustrated.
It is emphasized that the error correcting model in the embodiment of the present application can be end-to-end (sequence to
Sequence, seq2seq) deep learning model, can be generally divided into following type: Recognition with Recurrent Neural Network (Recurrent
Neural Networks, RNN) model, shot and long term memory network (Long Short-Term Memory, LSTM) model, gate
Cycling element (Gated Recurrent Unit, GRU) model and convolutional neural networks (Convolutional Neural
Network, CNN) model etc..Deep learning model is applied in error correction scene as error correcting model, relative to conventional error correction mould
The advantage of type (such as nGram model) is not needing to be manually error correcting model design rule.Deep learning model can utilize
To more information, for example, it is longer more information are merged etc. above, and more complicated rule can be learnt, relatively
It is stronger in the generalization ability of conventional error correction model, deep learning model, can only not only handle in training corpus probability of occurrence compared with
High situation.
When the type of used error correcting model is different, and the keyboard for being used for input coding is dummy keyboard, input bit
The screen area information set in sequence also can be different.For example, when error correcting model uses CNN model, in input position sequence
Screen area information may include screen area pictorial information when each coding in coded strings is entered;And work as error correcting model
When using RNN model, the input position sequence for inputting error correcting model may include when each coding in coded strings is entered
Screen area location information.That is, screen area information may include compiling in the corresponding input position sequence of coded strings
Screen area location information or screen area pictorial information and all codings when each coding in sequence is entered
The input sequence of screen area location information or screen area pictorial information.
Next, the screen area location information and screen area pictorial information when being entered to coding are introduced.
Firstly, being clicked when the screen area pictorial information of coding can be user and click the coding on the screen
The pictorial information of screen area, wherein the screen area pictorial information of the coding can reflect clicked screen area and exist
Location information in screen.For example, as shown in Fig. 2, working as user wants input coding " k ", and user uses finger in keyboard
When clicking key " K " in screen, the region of user's finger and the screen contact is region 202, it can by the region 202
The screen area clicked as user.At this point it is possible to regard the picture 201 of the coordinate information including the region 202 as the volume
The screen area pictorial information of code " k ", wherein the coordinate information in the region 202 is rectangular coordinate system of the region 102 in screen
Coordinate value in (i.e. horizontal axis is x-axis, and the longitudinal axis is y-axis).
Secondly, being clicked when the screen area location information of coding can be user and click the coding on the screen
The coordinate information of screen area.For example, the screen area location information of coding can be in the screen area that user is clicked
All pixels coordinate, it can including the coordinate of each pixel for being in contact with user's limbs in screen, wherein screen
All pixels coordinate in curtain region can be in the matrix form to indicate;As an example it is assumed that when user wants input coding
" n ", and 50 pixels when user clicks key " N " in screen using finger in keyboard, in user's finger and screen
It is in contact, then can first obtain this 50 pixels coordinate in the screen respectively, it then, can be by the seat of this 50 pixels
It is denoted as the screen area location information of the coding " n ".In another example the screen area location information of coding can be user institute point
The center position coordinates of the screen area hit, that is, coordinate of the central point of the screen area in screen;For example, false
If when user wants input coding " n ", and when user clicks key " N " in screen using finger in keyboard, user hand
Refer to that clicked screen area is circle, then can first obtain the coordinate of the circular center of circle in the screen, it then, can be with
Using coordinate of the center of circle in the screen as the screen area location information of the coding " n ".
As it can be seen that in the embodiment of the present application, since error correcting model is to be trained, learn according to history input position sequence
It obtains, artificially presets error-correction rule and the stiff problem of general rule in traditional approach in this way, avoiding, it can be efficient
To coded strings carry out error correction, when for complicated error correction situation, do not need as called in traditional approach a large amount of error-correction rule or
In order to save make-up time discard portion error-correction rule, not only reduce the make-up time, and can be improved error correction accuracy and
It is comprehensive.
With reference to the accompanying drawing, the various non-limiting embodiments of the application are described in detail.
Below with reference to exemplary application scene shown in Fig. 4, Fig. 3 is cooperated to illustrate provided by the application embodiment of the method
Input error correction method.It should be noted that application scenarios shown in Fig. 4 are merely for convenience of understanding the application and showing, this Shen
Embodiment please is unrestricted in this regard.
As shown in figure 3, this method can be with for a kind of method flow diagram for inputting error correction method provided by the embodiments of the present application
The following steps are included:
S301: the input position sequence of coded strings and the input coded strings is obtained.
After user is by keyboard (such as dummy keyboard in screen) input coding string, the coding can be first got
String and the corresponding screen area location information of each coding or screen area pictorial information in the coded strings;Example
Such as, when the user clicks a coding corresponding key in terminal device screen when, terminal device can detecte terminal device
A certain screen area in screen is contacted with user's limbs, and can by the screen area all pixels coordinate or
The center position coordinates of the screen area (or can be by the screen area as the corresponding screen area location information of the coding
Pictorial information as the corresponding screen area pictorial information of the coding).It then, can be by each coding in the coded strings point
Not corresponding screen area location information or screen area pictorial information and the corresponding whole screen areas position of the coded strings
The input sequence of confidence breath or screen area pictorial information, the input position sequence as the coded strings.
Next, with exemplary application scene shown in Fig. 4 illustration.It is assumed that when user wants to input in mobile phone
When Chinese character " today ", user by finger successively clicked in the dummy keyboard of mobile phone coding " j ", coding " u ", encode " n ",
It encodes " t ", coding " i ", coding " a " and encodes " n " respective corresponding key.Mobile phone can first get coded strings
When " juntian " and user click on key corresponding to each coding in the coded strings " juntian ", sent out with screen
The center position coordinates of the screen area of raw contact.And it is possible to by corresponding to each coding in the coded strings " juntian "
Input in center position coordinates and the coded strings " juntian " between center position coordinates corresponding to each coding is suitable
Sequence, as the corresponding input position sequence of coded strings " juntian ".Then, mobile phone can be right by the coded strings " juntian "
The input position sequence answered is sent to server.
S302: by the input position sequence inputting error correcting model;If judging the volume according to the input position sequence
There is wrong input coding in sequence, the error correction result of the coded strings is directed to by error correcting model output;Wherein, described
Error correcting model is obtained according to the training of history input position sequence.
It, can be by the corresponding input of the coded strings after getting coded strings and the corresponding input position sequence of coded strings
It is handled in position sequence input error correcting model.Then, the error-correction rule that error correcting model can learn in advance according to it, to this
Whole screen area location informations or screen area pictorial information in input position sequence are judged;If judging the input
Screen area location information or screen area pictorial information at least one mistake in position sequence, then can be for every
One wrong screen area location information or screen area pictorial information, according to the screen area location information of the mistake or
Person's screen area pictorial information determines in the coded strings and believes with the screen area location information of the mistake or screen area picture
Corresponding coding is ceased, it is thus possible to using the coding as having vicious input coding in the coded strings.For ease of description, can
It is referred to as wrong input coding will have vicious input coding in the coded strings.For example, it is assumed that user wants in mobile phone
It inputs Chinese character " I ", but is had input coded strings " wi " by keyboard, the coded strings as corresponding to Chinese character " I " are " wo ";Institute
The coding " i " in coded strings " wi " or coded strings " wi " can be used as to wrong input coding.
When error correcting model judges the input position according to input position sequence using its error-correction rule learnt in advance
When including the screen area location information or screen area pictorial information that at least one needs to carry out error correction in sequence
There is wrong input coding, so as to export the error correction result for being directed to the coded strings in coded strings to judge input.
Wherein, error correction result can be to obtained correct coding string after the progress error correction of wrong input coding.For example, false
If error correcting model according to the corresponding input position sequence of coded strings " wisho ", judges the coded strings in the coded strings " wisho "
" wi " there is wrong input coding " i " and coded strings " sho " to have wrong input coding " o ", then can be first defeated by the mistake
Enter coding " i " and be corrected as coding " o ", and the mistake input coding " o " is corrected as coding " i ";So as to obtain to mistake
Erroneous input encodes obtained correct coding string " woshi " after " i " and " o " progress error correction, which can be with
As the error correction result for being directed to the coded strings " wisho ".
Certainly, which can also be the corresponding candidate item of correct coding string, wherein the correct coding string can be
There are the coded strings of wrong input coding to carry out what error correction obtained this.Continue by taking above-mentioned coded strings " wisho " as an example, to mistake
After input coding " i " and " o " carry out error correction, the corresponding correct coding string " woshi " of the available coded strings " wisho ";It connects
, which can be used as and be directed to
The error correction result of the coded strings " wisho ".
Next, continuing by taking the exemplary application scene shown in Fig. 4 in S301 as an example.It is assumed that server shown in Fig. 4
In be stored with error correcting model, wherein the error correcting model be RNN model.Server receives the coded strings of mobile phone transmission
After " juntian " corresponding input position sequence, which can be somebody's turn to do
In error correcting model.The error-correction rule that the error correcting model can learn in advance according to it is somebody's turn to do to included by the input position sequence
Center position coordinates corresponding to each coding (i.e. screen area location information) are judged in coded strings " juntian ".By
Judge in center position coordinates included by the input position sequence in the error correcting model, the corresponding center of coding " u " is sat
The center position coordinates of mistake are designated as, so, the corresponding coding " u " of the center position coordinates of the mistake can be determined as the volume
Wrong input coding in sequence " juntian ".Then, which it is corresponding can to export the coded strings " juntian "
Error correction result, wherein can be by the corresponding correct coding string " jintian " of the coded strings " juntian " as the coding
The error correction result of " juntian " is gone here and there, alternatively, can be (such as " modern by the corresponding candidate item of correct coding string " jintian "
It ", " golden field ", " Jin Tian ", " Kintian " etc.) as the error correction result for being directed to the coded strings " juntian ".Finally, server can
It is sent so that the error correction result of the coded strings " juntian " will be directed to mobile phone.
S303: error correction is carried out for the coded strings according to the error correction result.
After the error correction result for coded strings for getting error correcting model output, can be directed to according to the error correction result should
Coded strings carry out error correction, and it is in other words user's displaying that the purpose of error correction, which is the candidate item for showing that user wishes to for user,
The user wants the upper candidate item shielded in S301 input coding string.When carrying out error correction for coded strings according to error correction result,
In order to realize prescribed fault-correction purpose, different error correcting systems can be used, such as can be directly to wrong input coding
Coded strings carry out error correction, to include the candidate shielded on user wishes in candidate item corresponding to the coded strings after error correction
, i.e., the correct coding string that user should input is obtained relative to the mistake by correcting encoder string, passes through correct coding string institute
Corresponding candidate item is to realize foregoing purpose;It is also possible to not carry out error correction to the coded strings with wrong input coding, but it is straight
Connect and error correction carried out to candidate item, by user wish on the candidate item shielded be added in the candidate item currently shown, equally also can be real
Existing foregoing purpose.
It can also be and carry out error correction for other information relevant to coded strings, herein not to the object for needing to carry out error correction
Do any restriction.
It should be noted that the error correction result of coded strings is different, according to error correction result coded strings are carried out with the mode of error correction
Also different.Next, how will introduce according to the different error correction results of same coded strings, error correction is carried out to coded strings.
As an example, when the error correction result for coded strings is to entangle to the wrong input coding in the coded strings
After mistake when obtained correct coding string, then the mode for carrying out error correction to the coded strings can be with are as follows: will entangle to the coded strings
Obtained correct coding string is shown after mistake.Continue by taking the exemplary application scene shown in Fig. 4 in S302 as an example, works as hand
Machine gets the error correction result for coded strings " juntian " as to obtained after the coded strings " juntian " progress error correction
When correct coding string " jintian ", as shown in figure 5, mobile phone can be by pinyin string shown in interface of input method " jun ' tian "
(i.e. coded strings " juntian ") replace with pinyin string " jin ' tian " (i.e. correct coding string " jintian ").
As an example, when the error correction result for coded strings is to carry out the obtained correct volume of error correction to the coded strings
When the corresponding candidate item of sequence, then the mode for carrying out error correction to the coded strings can be with are as follows: will the coded strings be carried out with institute after error correction
The obtained corresponding candidate item of correct coding string is shown.Continue with the exemplary application scene shown in Fig. 4 in S302 be
Example, when mobile phone gets the error correction result for coded strings " juntian " as to obtained by the coded strings " juntian " progress error correction
When correct coding string " jintian " the corresponding candidate item arrived, as shown in fig. 6, mobile phone can will be shown in interface of input method
The corresponding candidate item " an ancient unit of weight day " of pinyin string " jun ' tian " (i.e. coded strings " juntian "), " monarch day ", " army field ", " equal day " replace
Be changed to the corresponding candidate item " today " of pinyin string " jin ' tian " (i.e. correct coding string " jintian "), " golden field ", " Jin Tian ",
" Kintian ".
It is emphasized that being mentioned according to the mode that error correction result carries out error correction for coded strings in addition to above-mentioned in the present embodiment
Other than several possible modes arrived, it can also be other modes that may be implemented, no longer repeated herein.
As it can be seen that in the present embodiment, when detecting coding input behavior, obtaining coded strings and inputting the coded strings
Input position sequence, and by the input position sequence inputting into preset error correcting model.Since the error correcting model is that basis is gone through
History input position sequence is trained, therefore eliminates artificial setting rule by the error-correction rule that the error correcting model obtains
Influence, moreover, training the error correcting model be history input position sequence, the history input position sequence be user input compile
The position sequence pressed on the input when sequence can more embody personalized input feature vector when user's input.If
The error correcting model judges there is wrong input coding in the coded strings by input position sequence, can pass through the error correcting model
Output is directed to the coded strings error correction result, so as to carry out error correction for the coded strings according to the error correction result.In this way, due to
Error correcting model can more embody personalized input feature vector when user's input, so as to realize the personalization for being directed to user
Error correction, moreover, because the error-correction rule in error correcting model is to be trained according to history input position sequence, learn to obtain,
Avoid and artificially preset error-correction rule and the stiff problem of general rule in traditional approach, can efficiently to coded strings into
Row error correction when for complicated error correction situation, is not needed as called a large amount of error-correction rule in traditional approach or in order to save error correction
Time discard portion error-correction rule, not only reduces the make-up time, and can be improved error correction accuracy and comprehensive.
Since under different input conditions, the same user might have different input habits, so as to cause not
Same inputs situation by mistake, for this purpose, the mode of a variety of trained error correcting models is provided in the embodiment of the present application, for example, can use
History input position sequence that history input position sequence, different types of input keyboard under different input scenes are inputted or
Person's history input position sequence and historical time intervals sequence are trained error correcting model.Next, will be for wherein three kinds
The mode of training error correcting model is illustrated.
The mode of the first training error correcting model:
In this mode, will introduce how using the history input position sequence under different input scenes to error correcting model into
Row training.
Since user is under different input scenes, although inputting identical wrong input coding string by input method,
It is that the upper screen character that shields accords with corresponding correct coding string in selection is different, or the upper screen character Fu Buxiang shielded in selection
Together, however correct coding string corresponding to these different upper screen character symbols is identical.That is, the same mistake input
For coded strings under different input scenes, correct coding string corresponding to the mistake input coding string is different, or should
Correct coding string corresponding to mistake input coding string may be identical, but the upper screen character symbol shielded on selecting is not identical
's.In addition, the input habit of same user is also different due under different input scenes, cause user in different input fields
Under scape, it is desirable to which when inputting same coded strings, the input position sequence for actually entering coded strings is not identical.
Next, illustrating the same wrong input coding string not by taking wrong input coding string " humian " as an example
Under same input scene, correct coding string corresponding to the mistake input coding string is different.Assuming that user is in cuisines net
In standing, after input method mistake input coding string " humian ", selecting the upper screen character shielded symbol is " stewed noodles ", that is, in cuisines
Under the input scene of website, correct coding string corresponding to mistake input coding string " humian " is " huimian ";And the user
In search website, after input method mistake input coding string " humian ", selecting the upper screen character shielded symbol is " and face ",
That is, correct coding string corresponding to mistake input coding string " humian " is under the input scene of search website
"huimian";As it can be seen that respectively under the input scene of cuisines website and search website, although same input error input coding
The corresponding input position sequence of string " humian ", but corresponding to the corresponding input position sequence of input coding string " humian "
Correct coding string be different.
Next, illustrating the same wrong input coding string by taking wrong input coding string " xiagnjia " as an example and existing
Under different input scenes, correct coding string corresponding to the mistake input coding string may be identical, but shield on selecting
Upper screen character symbol be different.It is assumed that passing through input method input coding string when user is going up network mathematics guidance course
After " xiagnjia ", selecting the upper screen character shielded symbol is Chinese phrase " addition ", that is, teaches the input field of course in network mathematics
Under scape, correct coding string corresponding to mistake input coding string " xiagnjia " is " xiangjia ", and the correct coding string
" xiangjia " corresponding upper screen character symbol is " addition ";However, the user is by instant communication software and household with the shape of text
When formula is chatted, after input method input coding string " xiagnjia ", the upper screen character shielded symbol is selected " to think for Chinese phrase
Family ", that is, under the input scene of chat, correct coding string corresponding to mistake input coding string " xiagnjia " is
" xiangjia ", and the corresponding upper screen character symbol of the correct coding string " xiangjia " is " homesick ";As it can be seen that respectively in network mathematics
Under the input scene for teaching course and chat, although the corresponding input position of same input error input coding string " xiagnjia "
Sequence, but the symbol of upper screen character corresponding to the corresponding input position sequence of input coding string " xiagnjia " is different.
Next, being illustrated under different input scenes, it is desirable to input same coded strings by taking coded strings " ing " as an example
When, the input position sequence for actually entering coded strings is not identical.It is assumed that when user writes minutes in a meeting, due to needing
Other people speech is listened on one side, quickly cause the user to wish using input coding string in input normal direction tablet computer on one side
When input coding string " ing ", the corresponding screen area of coding " i " is often first clicked, then clicks the corresponding screen area of coding " g "
The corresponding screen area of coding " n " is then clicked in domain, so that the input position sequence that the tablet computer is got is coded strings
" ign " corresponding input position sequence: the corresponding screen area location information → corresponding screen area of coding " g " of coding " i "
The corresponding screen area location information of location information → coding " n ";And the user is in through mobile phone and friend with the shape of text
When formula is chatted, which wishes input coding string " ing ", then often first clicks and compile " i " corresponding screen area, then point
The corresponding screen area of coding " n " is hit, the corresponding screen area of coding " g " is then clicked, so that the input that the mobile phone is got
Position sequence is the corresponding input position sequence of coded strings " ing ": the corresponding screen area location information → coding of coding " i "
" n " the corresponding screen area location information → corresponding screen area location information of coding " g ";As it can be seen that writing meeting respectively
Under record and the input scene of chat, although practical institute's input coding string is corresponding it is desirable also to input coding string " ing "
Input position sequence is different.
So the error correction result that error correcting model is exported can meet use to can be realized under different input scenes
Family selects the habit of upper screen character symbol.In a kind of implementation of the embodiment of the present application, the error correcting model in S302 can be root
It is obtained according to the history input position sequence training under same input scene, can also include step before S302 correspondingly
It is rapid: obtain the corresponding target input scene of the coded strings, also, in S302 " by the input position sequence inputting error correction
Model " may include step: by the corresponding error correcting model of target input scene described in the input position sequence inputting.Wherein,
Scene when target input scene can be input coding string, such as can be chat scenario, map search scene, cuisines search
Scene, study scene etc..
It is understood that in the present embodiment, when obtaining coded strings and input the input position sequence of the coded strings
When, the corresponding target input scene of the coded strings can also be obtained.It is then possible to according to the target input scene, from it is each not
In the same corresponding error correcting model of input scene, the corresponding error correcting model of target input scene is determined.It then, can be defeated by this
Enter position sequence to input in the corresponding error correcting model of target input scene.In this way, the corresponding error correction mould of the target input scene
Type can be directed to input habit of the user under different input scenes, be sentenced according to the corresponding input position sequence of input coding string
There is wrong input coding in the coded strings of breaking, and export the error correction result for being directed to the coded strings, so as to efficiently to this
Coded strings carry out error correction.
The mode of second of trained error correcting model:
In this mode, the history input position sequence pair how introduction inputted using different types of input keyboard
Error correcting model is trained.
When user wants to input message by mobile phone, it usually needs by clicking the dummy keyboard input coding in mobile phone
String, wherein the dummy keyboard in mobile phone can be generally divided into full keyboard and nine grids keyboard two types.And due to same user
Input habit in different types of keyboard is different, and the user is caused to wish that input is same using different types of keyboard
When one coded strings, it is different for actually entering the input position sequence of coded strings.For example, being compiled due in nine grids keyboard
The position of the position key corresponding with coding " j " of the corresponding key of code " i " is adjacent, when user being caused to wish to input " i ", often
Mistake clicks the corresponding screen area of coding " j ", so, when the user wishes input coding string " yi " by nine grids keyboard
When, the corresponding screen area of coding " y " is often first clicked, the corresponding screen area of coding " j " is then clicked, so that mobile phone obtains
The input position sequence arrived is coded strings " yj " corresponding input position sequence: the corresponding screen area location information of coding " y "
The corresponding screen area location information of → coding " j ";And the position due to encoding " i " corresponding key in full keyboard and coding
" j " corresponding key positions remote from, so then will not when the user wishes input coding string " yi " using full keyboard
There is the case where mistake clicks other codings corresponding screen area, so that the input position sequence that mobile phone is got is coding
" yi " corresponding input position sequence.
So to enable error correcting model for input habit of the user in different types of keyboard, it can basis
The corresponding input position sequence of input coding string judges whether there is wrong input coding in the coded strings.In the embodiment of the present application
A kind of implementation in, the history that the error correcting model in S302 can be inputted according to by same type of input keyboard is defeated
Enter what position sequence training obtained, it correspondingly, can be comprising steps of obtaining the input position sequence pair before S302
The target keyboard type answered, also, " by the input position sequence inputting error correcting model " in S302 may include step: it will
The corresponding error correcting model of target keyboard type described in the input position sequence inputting.Wherein, target keyboard type can be defeated
Enter used keyboard type when coded strings, such as can be full keyboard, nine grids keyboard etc..
It is understood that in the present embodiment, when obtaining coded strings and input the input position sequence of the coded strings
When, the corresponding target keyboard type of the coded strings can also be obtained.It is then possible to according to the target keyboard type, from each key
In the corresponding error correcting model of disc-type, the corresponding error correcting model of target keyboard type is determined.It then, can be by the input position
In the sequence inputting corresponding error correcting model of target keyboard type.In this way, the corresponding error correcting model of target keyboard type
It, should according to the corresponding input position sequence judgement of input coding string to be directed to input habit of the user in different types of keyboard
There is wrong input coding in coded strings, and export the error correction result for being directed to the coded strings, so as to efficiently to the coding
String carries out error correction.
The mode of the third training error correcting model:
In this mode, how will introduce utilizes history input position sequence and historical time intervals sequence to error correcting model
Error correcting model is trained.
User is in the coding inputted in a syllable, between the input time in a syllable between adjacent two codings
It is very short every meeting, for example, encoding " n " when user inputs coding " n " and " i " in a syllable " ni " by keyboard and encoding
0.01s is divided between input time between " i ".And when user inputs the coded strings of two inter-syllables by keyboard, two sounds
The input time interval saved between corresponding input coding string can be long, such as, it is assumed that user is inputted by keyboard and is wrapped
When including coded strings " nihao " of syllable " ni " and syllable " hao ", user inputs syllable " ni " corresponding coded strings " ni " and input
0.1s is divided between input time between the corresponding coded strings " hao " of syllable " hao ".
So in order to improve error correcting model output for the accuracy of the error correction result of coded strings, in the embodiment of the present application
A kind of implementation in, error correcting model can be long according to the input time interval between two codings adjacent in coded strings
It is short, it determines in the corresponding input position sequence of the coded strings, the input position sequence of the corresponding same syllable, and root can be distinguished
According to the input position sequence of a corresponding syllable, judge that there is wrong input coding in the coded strings, and export and be directed to the coding
The error correction result of string.
Specifically, in the present embodiment the error correcting model in S302 can according to history input position sequence and when history
Between intervening sequence training obtain, correspondingly, S301 may include step: obtain the coded strings and the input coded strings
Input position sequence and time interval sequence, also, S302 may include step: by the input position sequence and it is described when
Between intervening sequence input error correcting model.Wherein, the time interval sequence of the coded strings may include arbitrary neighborhood in the coded strings
Two coding between input time interval and each time interval between record sequence.
It is understood that in the present embodiment, can first obtain coded strings and input the input position of the coded strings
Sequence and time interval sequence.It is then possible to which this is entangled by the input position sequence and the time interval sequence inputting error correcting model
Mismatch type can determine the input position sequence that the same syllable is corresponded in the input position sequence according to the time interval sequence
Column.For example, can determine the two volumes if the input time interval between two adjacent codings is less than time interval threshold value
The corresponding screen area location information of code or screen area pictorial information correspond to the same syllable, and can be by the two codings point
The input position sequence of not corresponding screen area location information or screen area pictorial information as the corresponding syllable;If
Input time interval between two adjacent coded strings is greater than time interval threshold value, then can determine that the two coded strings are distinguished
Corresponding input position sequence corresponds to different syllables, and can be using the corresponding input position sequence of a coded strings as correspondence
The input position sequence of one syllable, and using the corresponding input position sequence of another coded strings as another corresponding syllable
Input position sequence.Then, which can judge in the coded strings according to the input position sequence of the different syllables of correspondence
Whether there is wrong input coding, if so, the error correction result for the coded strings can be exported.
As an example it is assumed that time interval threshold value is 0.05s, and the coded strings got are " wengti " and the volume
The corresponding input position sequence of sequence " wengti " and time interval sequence, wherein the time interval sequence are as follows: coding " w " with
Encode input time interval 0.01s → coding between the input time interval 0.01s → coding " e " and coding " n " between " e "
" n " and coding " g " between input time interval 0.01s → coding " g " with encode " t " between input time interval 0.1s →
It encodes " t " and encodes the input time interval 0.01s between " i ".By the corresponding input position sequence of coded strings " wengti "
After column and time interval sequence inputting error correcting model, which can be first according to the time interval sequence and time interval threshold
Value 0.05s is determined in the corresponding input position sequence of coded strings " wengti ", corresponds to the input position sequence of the same syllable
Column.Specifically, due to coding " w " and coding " e " between input time interval 0.01s, coding " e ", coding " n " between it is defeated
Angle of incidence interval 0.01s and coding " n " and the input time interval 0.01s encoded between " g " are respectively less than time interval threshold value
0.05s, and the input time interval 0.1s for encoding " g " and encoding between " t " is greater than time interval threshold value 0.05s, so, it can be with
Determine that the corresponding screen area location information of coding " w ", " e ", " n " and " g " or screen area pictorial information are corresponding same
One syllable A, and can be by the corresponding screen area location information of coding " w ", " e ", " n " and " g " or screen area figure
Input position sequence of the piece information as corresponding syllable A.Due to the input time interval between coding " t " and coding " i "
0.01s is less than time interval threshold value 0.05s, therefore, it is determined coding " t " and coding corresponding screen area position " i "
Confidence breath or screen area pictorial information correspond to the same syllable B, and can be corresponding by coding " t " and coding " i "
The input position sequence of screen area location information or screen area pictorial information as corresponding syllable B.Then, the error correction mould
Type can judge have mistake defeated in the coded strings " wengti " according to the input position sequence for respectively corresponding syllable A and syllable B
Enter coding, and the error correction result for the coded strings " wengti " can be exported.
As it can be seen that by obtaining the corresponding input position sequence of input coding string and time interval sequence, when can be according to this
Between intervening sequence, determine in the corresponding input position sequence of the coded strings, the input position sequence of the corresponding same syllable.This
Sample, error correcting model can be according to the input position sequences of the different syllables of correspondence, to the wrong input coding in the input coding string
More accurately judged, so as to improve the accuracy for the error correction result that the error correcting model exports.
It should be noted that the mode of three of the above training error correcting model can also be tied with stand-alone training error correcting model
It closes and states the combinations of any various ways error correcting model is trained.
After the mode for having introduced a variety of trained error correcting models in the embodiment of the present application, it will introduce real corresponding to Fig. 3
On the basis of applying example, when " error correcting model " in S302 can not only export the error correction result for coded strings, it can also export
When the corresponding confidence level score of the error correction result, how according to the confidence level score of the error correction result and the error correction result for compiling
Sequence carries out error correction.
When user is using input method input coding string in device end, due to the key in device end screen compared with
It is small, when user rapidly inputs coded strings, be often difficult to accurately to click user and wish the key clicked, but clicked with
User wishes other adjacent keys of the key clicked.For example, due in full keyboard, the corresponding key of coding " i " respectively with
It is adjacent to encode " u " corresponding key, the corresponding key of coding " o ";When user wishes input coding " u ", often due to input
It is too fast, the corresponding key of coding " i " is mistakenly clicked, it is often wrong when so that the user being caused to wish input coding string " yu "
Erroneous input coded strings " yi ";And when the user wishes input coding " o ", it is too fast due to inputting once in a while, mistakenly click
" i " corresponding key is encoded, when the user being caused to wish input coding string " yo ", mistake has input coded strings " yi " once in a while.This
Sample, when user passes through input method input coding string " yi ", user is practical to wish that the coded strings inputted are " yu " a possibility that ratio
Coded strings " yo " are bigger.
So in order to determine the biggish error correction result of credibility from the error correction result that error correcting model exports, and
Error correction can be carried out for coded strings according to the error correction result, to wish that the corresponding candidate item of the coded strings after error correction may include
User wishes the content of input.In a kind of implementation of the embodiment of the present application, " pass through the error correcting model in S302
Output is directed to the error correction result of the coded strings " may include step: the coded strings are directed to by error correcting model output
Error correction result and the corresponding confidence level score of the error correction result;Wherein, the confidence level score is for indicating the error correction
As a result credibility.Correspondingly, S303 may include step: if the confidence level score of the error correction result is greater than threshold value, root
According to the confidence level score of the error correction result and the error correction result, error correction is carried out for the coded strings.
Wherein, the corresponding confidence level score of error correction result can be used to indicate that the credibility of the error correction result.In one kind
In example, the numberical range of confidence level score can be 0 to 100.It is understood that the corresponding confidence level score of error correction result
Higher, then the confidence level degree of the error correction result is higher, i.e., the error correction result meet user select upper screen character symbol habit can
Energy property is bigger, conversely, then the confidence level degree of the error correction result is lower, i.e., the error correction result meets user's selection upper screen character symbol
A possibility that habit, is smaller.
It should be noted that in coded strings any two coding between input time interval it is shorter, it can be said that bright
Input speed of the user when inputting the two codings is faster, occurs to click when inputting the two codings so as to cause user and press
The probability of key error is bigger.That is, the input time interval between any two coding in coded strings is shorter, then the volume
Have the probability of wrong input coding bigger in sequence, error correcting model output is corresponding credible for the error correction result of the coded strings
Degree point is also higher;Conversely, then having the probability of wrong input coding smaller in the coded strings, error correcting model output is directed to the volume
The corresponding confidence level point of the error correction result of sequence is also lower.
In the present embodiment, the error correction result for being directed to coded strings by error correcting model output is corresponding with the error correction result can
After confidence score, it can first judge whether the corresponding confidence level score of the error correction result is greater than threshold value.If the error correction result is corresponding
Confidence level score be greater than threshold value, then illustrate the error correction result meet user select upper screen character symbol habit, can be entangled according to this
The confidence level score of wrong result and the error correction result carries out error correction for coded strings, conversely, then illustrating that the error correction result is not met
User selects the habit of upper screen character symbol, it is not possible to according to the confidence level score of the error correction result and the error correction result, to coded strings
Carry out error correction.
Next, will introduce when the corresponding confidence level score of error correction result is greater than threshold value, how according to error correction result and
The confidence level score of the error correction result carries out a variety of implementations of error correction for coded strings:
In one implementation, when for coded strings error correction result be to the wrong input coding in the coded strings into
After row error correction when obtained correct coding string, the mode for carrying out error correction to the coded strings can be with are as follows: first by each correct coding
Corresponding confidence level score of going here and there carries out ranking from high to low;It is then possible to determine the highest confidence level of ranking from the ranking
Score or N number of confidence level score in the top (N is the positive integer greater than 1), and can will divide with the confidence level determined
The corresponding correct coding string of number is shown.
In one implementation, when the error correction result for coded strings is obtained just to coded strings progress error correction
When the corresponding candidate item of true coded strings, the mode for carrying out error correction to the coded strings can be with are as follows: first corresponding according to each candidate item
Confidence level score determines the display location of each candidate item;Then, according to the display location, each candidate item is shown.General feelings
Under condition, before the display location of the higher candidate item of confidence level score being placed on the lower candidate item of confidence level score, this
In repeat no more.
As it can be seen that in the embodiment of the present application, by error correcting model output for the error correction result and the error correction knot of coded strings
The corresponding confidence level score of fruit can be according to the error correction result if the confidence level score of the error correction result is greater than threshold value
With the confidence level score of the error correction result, error correction is carried out for the coded strings.In this way, can be entangled from what error correcting model exported
The biggish error correction result of credibility is determined in wrong result, and error correction can be carried out for coded strings according to the error correction result, is made
The corresponding candidate item of coded strings after obtaining error correction may include the content that user wishes input.
It should be noted that being so that error correcting model output is more in line with the individual of user for the error correction result of coded strings
Habit.In a kind of implementation of the embodiment of the present application, in the confidence level score needle according to error correction result and the error correction result
Error correction is carried out to coded strings, and the corresponding correct coding string of the coded strings or the corresponding candidate item of the correct coding string are carried out
After displaying, if user has selected the candidate item of upper screen in the corresponding candidate item of correct coding string, it is possible to by the volume
The corresponding input position sequence of sequence can use the history input position sequence to error correction as history input position sequence
Model is trained.In this way, can so that the error correction result based on error correcting model output is more in line with the personal habits of user,
So that it is more accurate to the coded strings progress obtained correct coding string of error correction according to the error correction result, improve user's body
It tests.
Referring to Fig. 7, showing a kind of input error correction device, described device in the embodiment of the present application includes: acquiring unit
701, input unit 702, output unit 703 and error correction unit 704;
The acquiring unit 701, for obtaining the input position sequence of coded strings and the input coded strings;
The input unit 702, for by the input position sequence inputting error correcting model, according to the error correcting model
The training of history input position sequence obtains;
The output unit 703, if the error correcting model is defeated for judging there is wrong input coding in the coded strings
It is directed to the error correction result of the coded strings out;
The error correction unit 704, for carrying out error correction for the coded strings according to the error correction result.
Optionally, the history input position sequence includes that the corresponding input position of coded strings is correctly entered in historical data
Sequence input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched volume of character in history with corresponding in history input coding string
Sequence.
Optionally, the input position sequence includes input bit confidence when each coding in the coded strings is entered
Breath;The input position information includes screen area information or key information.
Optionally, the screen area information includes screen area location information or screen area pictorial information;It is described
The region that screen area is clicked when clicking coding on the screen by user;Wherein, the screen area location information includes institute
State the center position coordinates of all pixels coordinate or the screen area in screen area.
Optionally, the key information includes the key that user is clicked when input coding in physical keyboard.
Optionally, the error correcting model is to be obtained according to the combined training of any one or more in following manner:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;
The acquiring unit 701 is also used to obtain the corresponding target input scene of the coded strings;
The input unit 702 is also used to entangle target input scene is corresponding described in the input position sequence inputting
Mismatch type;
The second way are as follows:
The history input position sequence that is inputted according to the error correcting model by same type of input keyboard is trained
It arrives;
The acquiring unit 701 is also used to obtain the corresponding target keyboard type of the input position sequence;
The input unit 702 is also used to entangle target keyboard type is corresponding described in the input position sequence inputting
Mismatch type;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;
The acquiring unit 701 is also used to obtain the input position sequence of the coded strings and the input coded strings
With time interval sequence;
The input unit 702 is also used to the input position sequence and the time interval sequence inputting error correction mould
Type.
Optionally, the output unit 703 is also used to export the error correction for being directed to the coded strings by the error correcting model
As a result the corresponding confidence level score with the error correction result;Wherein, the confidence level score is for indicating the error correction result
Credibility;
The error correction unit 704 is entangled according to if the confidence level score for being also used to the error correction result is greater than threshold value
The confidence level score of wrong result and the error correction result carries out error correction for the coded strings.
Referring to Fig. 8, input error correction apparatus 800 may include following one or more components: processing component 802, memory
804, power supply module 806, multimedia component 806, audio component 810, the interface 812 of input/output (I/O), sensor module
814 and communication component 816.
Processing component 802 usually control equipment 800 integrated operation, such as with display, telephone call, data communication, phase
Machine operation and record operate associated operation.Processing component 802 may include that one or more processors 820 refer to execute
It enables, to perform all or part of the steps of the methods described above.In addition, processing component 802 may include one or more modules, just
Interaction between processing component 802 and other assemblies.For example, processing component 802 may include multi-media module, it is more to facilitate
Interaction between media component 806 and processing component 802.
Memory 804 is configured as storing various types of data to support the operation in equipment 800.These data are shown
Example includes the instruction of any application or method for operating in equipment 800, contact data, and telephone book data disappears
Breath, picture, video etc..Memory 804 can be by any kind of volatibility or non-volatile memory device or their group
It closes and realizes, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM) is erasable to compile
Journey read-only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash
Device, disk or CD.
Power supply module 806 provides electric power for the various assemblies of equipment 800.Power supply module 806 may include power management system
System, one or more power supplys and other with for equipment 800 generate, manage, and distribute the associated component of electric power.
Multimedia component 806 includes the screen of one output interface of offer between the equipment 800 and user.One
In a little embodiments, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen
Curtain may be implemented as touch screen, to receive input signal from the user.Touch panel includes one or more touch sensings
Device is to sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense touch or sliding action
Boundary, but also detect duration and pressure associated with the touch or slide operation.In some embodiments, more matchmakers
Body component 806 includes a front camera and/or rear camera.When equipment 800 is in operation mode, such as screening-mode or
When video mode, front camera and/or rear camera can receive external multi-medium data.Each front camera and
Rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 810 is configured as output and/or input audio signal.For example, audio component 810 includes a Mike
Wind (MIC), when equipment 800 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone is matched
It is set to reception external audio signal.The received audio signal can be further stored in memory 804 or via communication set
Part 816 is sent.In some embodiments, audio component 810 further includes a loudspeaker, is used for output audio signal.
I/O interface 812 provides interface between processing component 802 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include, but are not limited to: home button, volume button, start button and lock
Determine button.
Sensor module 814 includes one or more sensors, and the state for providing various aspects for equipment 800 is commented
Estimate.Such as sensor module 814 can detecte the state that opens/closes of equipment 800, the relative positioning of component, such as described
Component is the display and keypad of equipment 800, and sensor module 814 can be with 800 1 components of detection device 800 or equipment
Position change, the existence or non-existence that user contacts with equipment 800,800 orientation of equipment or acceleration/deceleration and equipment 800
Temperature change.Sensor module 814 may include proximity sensor, be configured to detect without any physical contact
Presence of nearby objects.Sensor module 814 can also include optical sensor, such as CMOS or ccd image sensor, at
As being used in application.In some embodiments, which can also include acceleration transducer, gyro sensors
Device, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 816 is configured to facilitate the communication of wired or wireless way between equipment 800 and other equipment.Equipment
800 can access the wireless network based on communication standard, such as WiFi, 2G or 3G or their combination.In an exemplary implementation
In example, communication component 816 receives broadcast singal or broadcast related information from external broadcasting management system via broadcast channel.
In one exemplary embodiment, the communication component 816 further includes near-field communication (NFC) module, to promote short range communication.Example
Such as, NFC module can be based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band (UWB) technology,
Bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, equipment 800 can be believed by one or more application specific integrated circuit (ASIC), number
Number processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing the above method.
The embodiment of the present application provides a kind of input error correction apparatus.The equipment includes memory and one or one
A above program, perhaps more than one program is stored in memory and is configured to by one or one for one of them
It includes the instruction for performing the following operation that the above processor, which executes the one or more programs:
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is according to the training of history input position sequence
It obtains;
If judging that there is wrong input coding in the coded strings, error correcting model output the entangling for the coded strings
Wrong result;
Error correction is carried out for the coded strings according to the error correction result.
Optionally, the history input position sequence includes that the corresponding input position of coded strings is correctly entered in historical data
Sequence input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched volume of character in history with corresponding in history input coding string
Sequence.
Optionally, the input position sequence includes input bit confidence when each coding in the coded strings is entered
Breath;The input position information includes screen area information or key information.
Optionally, the screen area information includes screen area location information or screen area pictorial information;It is described
The region that screen area is clicked when clicking coding on the screen by user;Wherein, the screen area location information includes institute
State the center position coordinates of all pixels coordinate or the screen area in screen area.
Optionally, the key information includes the key that user is clicked when input coding in physical keyboard.
Optionally, the error correcting model is to be obtained according to the combined training of any one or more in following manner:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;It is incited somebody to action described
Before the input position sequence inputting error correcting model, further includes:
Obtain the corresponding target input scene of the coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target input scene described in the input position sequence inputting;
The second way are as follows:
The history input position sequence that is inputted according to the error correcting model by same type of input keyboard is trained
It arrives;It is described by the input position sequence inputting error correcting model before, further includes:
Obtain the corresponding target keyboard type of the input position sequence;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target keyboard type described in the input position sequence inputting;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;It is described to obtain
Take the input position sequence of coded strings and the input coded strings, comprising:
Obtain the input position sequence and time interval sequence of the coded strings and the input coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the input position sequence and the time interval sequence inputting error correcting model.
Optionally, the error correction result that the coded strings are directed to by error correcting model output, comprising:
It is corresponding credible for the error correction result of the coded strings and the error correction result by error correcting model output
Spend score;Wherein, the confidence level score is used to indicate the credibility of the error correction result;
It is described that error correction is carried out for the coded strings according to the error correction result, comprising:
If the confidence level score of the error correction result is greater than threshold value, according to the error correction result and the error correction result
Confidence level score carries out error correction for the coded strings.
The embodiment of the present application also provides a kind of non-transitorycomputer readable storage medium including instruction, for example including
The memory 804 of instruction, above-metioned instruction can be executed by the processor 820 of device 800 to complete the above method.For example, described non-
Provisional computer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, floppy disk and light number
According to storage equipment etc..
A kind of non-transitorycomputer readable storage medium, when the instruction in the storage medium is by the processing of electronic equipment
When device executes, so that electronic equipment is able to carry out a kind of input error correction method, which comprises
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is according to the training of history input position sequence
It obtains;
If judging that there is wrong input coding in the coded strings, error correcting model output the entangling for the coded strings
Wrong result;
Error correction is carried out for the coded strings according to the error correction result.
Optionally, the history input position sequence includes that the corresponding input position of coded strings is correctly entered in historical data
Sequence input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched volume of character in history with corresponding in history input coding string
Sequence.
Optionally, the input position sequence includes input bit confidence when each coding in the coded strings is entered
Breath;The input position information includes screen area information or key information.
Optionally, the screen area information includes screen area location information or screen area pictorial information;It is described
The region that screen area is clicked when clicking coding on the screen by user;Wherein, the screen area location information includes institute
State the center position coordinates of all pixels coordinate or the screen area in screen area.
Optionally, the key information includes the key that user is clicked when input coding in physical keyboard.
Optionally, the error correcting model is to be obtained according to the combined training of any one or more in following manner:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;It is incited somebody to action described
Before the input position sequence inputting error correcting model, further includes:
Obtain the corresponding target input scene of the coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target input scene described in the input position sequence inputting;
The second way are as follows:
The history input position sequence that is inputted according to the error correcting model by same type of input keyboard is trained
It arrives;It is described by the input position sequence inputting error correcting model before, further includes:
Obtain the corresponding target keyboard type of the input position sequence;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target keyboard type described in the input position sequence inputting;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;It is described to obtain
Take the input position sequence of coded strings and the input coded strings, comprising:
Obtain the input position sequence and time interval sequence of the coded strings and the input coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the input position sequence and the time interval sequence inputting error correcting model.
Optionally, the error correction result that the coded strings are directed to by error correcting model output, comprising:
It is corresponding credible for the error correction result of the coded strings and the error correction result by error correcting model output
Spend score;Wherein, the confidence level score is used to indicate the credibility of the error correction result;
It is described that error correction is carried out for the coded strings according to the error correction result, comprising:
If the confidence level score of the error correction result is greater than threshold value, according to the error correction result and the error correction result
Confidence level score carries out error correction for the coded strings.
Fig. 9 is the structural schematic diagram of server in the embodiment of the present application.The server 900 can be due to configuration or performance be different
Generate bigger difference, may include one or more central processing units (central processing units,
CPU) 922 (for example, one or more processors) and memory 932, one or more storage application programs 942 or
The storage medium 930 (such as one or more mass memory units) of data 944.Wherein, memory 932 and storage medium
930 can be of short duration storage or persistent storage.The program for being stored in storage medium 930 may include one or more modules
(diagram does not mark), each module may include to the series of instructions operation in server.Further, central processing unit
922 can be set to communicate with storage medium 930, and the series of instructions behaviour in storage medium 930 is executed on server 900
Make.
Server 900 can also include one or more power supplys 926, one or more wired or wireless networks
Interface 950, one or more input/output interfaces 958, one or more keyboards 956, and/or, one or one
The above operating system 941, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM etc..
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the application
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the application, these modifications, purposes or
Person's adaptive change follows the general principle of the application and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the application are by following
Claim is pointed out.
It should be understood that the application is not limited to the precise structure that has been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.Scope of the present application is only limited by the accompanying claims
The foregoing is merely the preferred embodiments of the application, not to limit the application, it is all in spirit herein and
Within principle, any modification, equivalent replacement, improvement and so on be should be included within the scope of protection of this application.
Claims (10)
1. a kind of input error correction method, which is characterized in that the described method includes:
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is to be obtained according to the training of history input position sequence
's;
If judging there is wrong input coding in the coded strings, the error correcting model output is directed to the error correction knot of the coded strings
Fruit;
Error correction is carried out for the coded strings according to the error correction result.
2. the method according to claim 1, wherein the history input position sequence include in historical data just
The really corresponding input position sequence of input coding string input position sequence corresponding with wrong input coding string;
Wherein, the wrong input coding string is to shield the unmatched coding of character in history with corresponding in history input coding string
String.
3. the method according to claim 1, wherein the input position sequence includes each in the coded strings
Input position information when a coding is entered;The input position information includes screen area information or key information.
4. according to the method described in claim 3, it is characterized in that, the screen area information includes screen area location information
Or screen area pictorial information;The region that the screen area is clicked when clicking coding on the screen by user;Wherein, institute
State the center that screen area location information includes all pixels coordinate or the screen area in the screen area
Coordinate.
5. according to the method described in claim 3, it is characterized in that, the key information includes that user inputs in physical keyboard
The key clicked when coding.
6. according to claim 1 to any method in 5, which is characterized in that the error correcting model is according to following manner
In any one or more combined training obtain:
First way are as follows:
The error correcting model is to be obtained according to the history input position sequence training under same input scene;It is described will be described
Before input position sequence inputting error correcting model, further includes:
Obtain the corresponding target input scene of the coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target input scene described in the input position sequence inputting;
The second way are as follows:
What the history input position sequence training inputted according to the error correcting model by same type of input keyboard obtained;
It is described by the input position sequence inputting error correcting model before, further includes:
Obtain the corresponding target keyboard type of the input position sequence;
It is described by the input position sequence inputting error correcting model, comprising:
By the corresponding error correcting model of target keyboard type described in the input position sequence inputting;
The third mode are as follows:
The error correcting model is obtained according to history input position sequence and the training of historical time intervals sequence;The acquisition is compiled
The input position sequence of sequence and the input coded strings, comprising:
Obtain the input position sequence and time interval sequence of the coded strings and the input coded strings;
It is described by the input position sequence inputting error correcting model, comprising:
By the input position sequence and the time interval sequence inputting error correcting model.
7. according to claim 1 to any method in 5, which is characterized in that described to export needle by the error correcting model
To the error correction result of the coded strings, comprising:
By error correcting model output for the error correction result of the coded strings and the corresponding confidence level point of the error correction result
Number;Wherein, the confidence level score is used to indicate the credibility of the error correction result;
It is described that error correction is carried out for the coded strings according to the error correction result, comprising:
If the confidence level score of the error correction result is greater than threshold value, according to the credible of the error correction result and the error correction result
Score is spent, carries out error correction for the coded strings.
8. a kind of input error correction device, which is characterized in that described device includes: acquiring unit, input unit, output unit and entangles
Wrong unit;
The acquiring unit, for obtaining the input position sequence of coded strings and the input coded strings;
The input unit, for by the input position sequence inputting error correcting model, the error correcting model to be defeated according to history
Enter what position sequence training obtained;
The output unit, if the error correcting model output is directed to for judging there is wrong input coding in the coded strings
The error correction result of the coded strings;
The error correction unit, for carrying out error correction for the coded strings according to the error correction result.
9. a kind of input error correction apparatus, which is characterized in that the equipment includes memory and one or more than one
Program, perhaps more than one program is stored in memory and is configured to by one or more than one processing for one of them
It includes the instruction for performing the following operation that device, which executes the one or more programs:
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is to be obtained according to the training of history input position sequence
's;
If judging there is wrong input coding in the coded strings, the error correcting model output is directed to the error correction knot of the coded strings
Fruit;
Error correction is carried out for the coded strings according to the error correction result.
10. a kind of non-transitorycomputer readable storage medium, which is characterized in that when the instruction in the storage medium is by electronics
When the processor of equipment executes, so that electronic equipment is able to carry out a kind of input error correction method, which comprises
Obtain the input position sequence of coded strings and the input coded strings;
By the input position sequence inputting error correcting model, the error correcting model is to be obtained according to the training of history input position sequence
's;
If judging there is wrong input coding in the coded strings, the error correcting model output is directed to the error correction knot of the coded strings
Fruit;
Error correction is carried out for the coded strings according to the error correction result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810339438.0A CN110389666B (en) | 2018-04-16 | 2018-04-16 | Input error correction method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810339438.0A CN110389666B (en) | 2018-04-16 | 2018-04-16 | Input error correction method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110389666A true CN110389666A (en) | 2019-10-29 |
CN110389666B CN110389666B (en) | 2024-07-12 |
Family
ID=68283027
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810339438.0A Active CN110389666B (en) | 2018-04-16 | 2018-04-16 | Input error correction method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110389666B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112905103A (en) * | 2021-03-05 | 2021-06-04 | 北京小米移动软件有限公司 | False touch processing method and device and storage medium |
CN114004320A (en) * | 2021-10-15 | 2022-02-01 | 库卡机器人(广东)有限公司 | Graphic code generation method and device, graphic code identification method and device, and electronic equipment |
CN115509371A (en) * | 2021-06-22 | 2022-12-23 | 北京搜狗科技发展有限公司 | Key identification method and device for identifying keys |
Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101241514A (en) * | 2008-03-21 | 2008-08-13 | 北京搜狗科技发展有限公司 | Method for creating error-correcting database, automatic error correcting method and system |
CN101710262A (en) * | 2009-12-11 | 2010-05-19 | 北京搜狗科技发展有限公司 | Error correction method and error correction device of characters |
CN102135814A (en) * | 2011-03-30 | 2011-07-27 | 北京搜狗科技发展有限公司 | Word input method and system |
US8209183B1 (en) * | 2011-07-07 | 2012-06-26 | Google Inc. | Systems and methods for correction of text from different input types, sources, and contexts |
CN103389915A (en) * | 2013-07-23 | 2013-11-13 | 百度在线网络技术(北京)有限公司 | Input error correcting method, input error correcting device, input error correcting server and input error correcting server system |
US20140104175A1 (en) * | 2012-10-16 | 2014-04-17 | Google Inc. | Feature-based autocorrection |
CN104246663A (en) * | 2013-12-31 | 2014-12-24 | 华为终端有限公司 | Character string input control method and device |
CN104915264A (en) * | 2015-05-29 | 2015-09-16 | 北京搜狗科技发展有限公司 | Input error-correction method and device |
CN105302336A (en) * | 2015-10-30 | 2016-02-03 | 北京搜狗科技发展有限公司 | Input error correction method and apparatus |
CN105491174A (en) * | 2014-09-18 | 2016-04-13 | 中国电信股份有限公司 | Domain name error correction method, server and system |
CN106484132A (en) * | 2015-09-02 | 2017-03-08 | 北京搜狗科技发展有限公司 | A kind of input error correction method and input subtraction unit |
CN106484131A (en) * | 2015-09-02 | 2017-03-08 | 北京搜狗科技发展有限公司 | A kind of input error correction method and input subtraction unit |
CN107102746A (en) * | 2016-02-19 | 2017-08-29 | 北京搜狗科技发展有限公司 | Candidate word generation method, device and the device generated for candidate word |
CN107229348A (en) * | 2016-03-23 | 2017-10-03 | 北京搜狗科技发展有限公司 | A kind of input error correction method, device and the device for inputting error correction |
CN107436691A (en) * | 2016-05-26 | 2017-12-05 | 北京搜狗科技发展有限公司 | A kind of input method carries out method, client, server and the device of error correction |
CN107544684A (en) * | 2016-06-29 | 2018-01-05 | 北京搜狗科技发展有限公司 | A kind of candidate word display methods and device |
CN107870677A (en) * | 2016-09-23 | 2018-04-03 | 北京搜狗科技发展有限公司 | A kind of input method, device and the device for input |
-
2018
- 2018-04-16 CN CN201810339438.0A patent/CN110389666B/en active Active
Patent Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101241514A (en) * | 2008-03-21 | 2008-08-13 | 北京搜狗科技发展有限公司 | Method for creating error-correcting database, automatic error correcting method and system |
CN101710262A (en) * | 2009-12-11 | 2010-05-19 | 北京搜狗科技发展有限公司 | Error correction method and error correction device of characters |
CN102135814A (en) * | 2011-03-30 | 2011-07-27 | 北京搜狗科技发展有限公司 | Word input method and system |
US8209183B1 (en) * | 2011-07-07 | 2012-06-26 | Google Inc. | Systems and methods for correction of text from different input types, sources, and contexts |
US20140104175A1 (en) * | 2012-10-16 | 2014-04-17 | Google Inc. | Feature-based autocorrection |
CN103389915A (en) * | 2013-07-23 | 2013-11-13 | 百度在线网络技术(北京)有限公司 | Input error correcting method, input error correcting device, input error correcting server and input error correcting server system |
CN104246663A (en) * | 2013-12-31 | 2014-12-24 | 华为终端有限公司 | Character string input control method and device |
CN105491174A (en) * | 2014-09-18 | 2016-04-13 | 中国电信股份有限公司 | Domain name error correction method, server and system |
CN104915264A (en) * | 2015-05-29 | 2015-09-16 | 北京搜狗科技发展有限公司 | Input error-correction method and device |
CN106484131A (en) * | 2015-09-02 | 2017-03-08 | 北京搜狗科技发展有限公司 | A kind of input error correction method and input subtraction unit |
CN106484132A (en) * | 2015-09-02 | 2017-03-08 | 北京搜狗科技发展有限公司 | A kind of input error correction method and input subtraction unit |
CN105302336A (en) * | 2015-10-30 | 2016-02-03 | 北京搜狗科技发展有限公司 | Input error correction method and apparatus |
CN107102746A (en) * | 2016-02-19 | 2017-08-29 | 北京搜狗科技发展有限公司 | Candidate word generation method, device and the device generated for candidate word |
CN107229348A (en) * | 2016-03-23 | 2017-10-03 | 北京搜狗科技发展有限公司 | A kind of input error correction method, device and the device for inputting error correction |
CN107436691A (en) * | 2016-05-26 | 2017-12-05 | 北京搜狗科技发展有限公司 | A kind of input method carries out method, client, server and the device of error correction |
CN107544684A (en) * | 2016-06-29 | 2018-01-05 | 北京搜狗科技发展有限公司 | A kind of candidate word display methods and device |
CN107870677A (en) * | 2016-09-23 | 2018-04-03 | 北京搜狗科技发展有限公司 | A kind of input method, device and the device for input |
Non-Patent Citations (4)
Title |
---|
HE SHAOFANG等: "Automatic syllable segmentation algorithm of Chinese speech based on MF-DFA", 《SPEECH COMMUNICATION》, vol. 92, 30 September 2017 (2017-09-30), pages 42 - 51, XP085143048, DOI: 10.1016/j.specom.2017.04.003 * |
SHA SHA等: "Automatic Chinese Topic Term Spelling Correction in Online Pinyin Input", 《PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON HUMAN-CENTRIC COMPUTING 2011 AND EMBEDDED AND MULTIMEDIA COMPUTING 2011》, 1 January 2011 (2011-01-01), pages 23 * |
刘磊: "搜索引擎缓存替换与容错检索技术研究", 《万方数据库》, 31 December 2013 (2013-12-31), pages 8 - 45 * |
谢旭东等: "一个基于混合语言模型的日文识别后处理系统", 《计算机工程与应用》, vol. 38, no. 14, 31 July 2002 (2002-07-31), pages 68 - 72 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112905103A (en) * | 2021-03-05 | 2021-06-04 | 北京小米移动软件有限公司 | False touch processing method and device and storage medium |
CN112905103B (en) * | 2021-03-05 | 2022-07-22 | 北京小米移动软件有限公司 | False touch processing method and device and storage medium |
CN115509371A (en) * | 2021-06-22 | 2022-12-23 | 北京搜狗科技发展有限公司 | Key identification method and device for identifying keys |
CN114004320A (en) * | 2021-10-15 | 2022-02-01 | 库卡机器人(广东)有限公司 | Graphic code generation method and device, graphic code identification method and device, and electronic equipment |
CN114004320B (en) * | 2021-10-15 | 2023-11-28 | 库卡机器人(广东)有限公司 | Method and device for generating graphic code, method and device for identifying graphic code and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN110389666B (en) | 2024-07-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102270394B1 (en) | Method, terminal, and storage medium for recognizing an image | |
WO2021077529A1 (en) | Neural network model compressing method, corpus translation method and device thereof | |
WO2016165557A1 (en) | Method and device for realizing verification code | |
CN109871896A (en) | Data classification method, device, electronic equipment and storage medium | |
CN107436691A (en) | A kind of input method carries out method, client, server and the device of error correction | |
US10303340B2 (en) | Method, device and system for user interface adaptation | |
CN107527059A (en) | Character recognition method, device and terminal | |
CN103529934A (en) | Method and apparatus for processing multiple inputs | |
CN107221330A (en) | Punctuate adding method and device, the device added for punctuate | |
CN108345581A (en) | A kind of information identifying method, device and terminal device | |
CN107845062A (en) | image generating method and device | |
CN110389666A (en) | A kind of input error correction method and device | |
CN108958503A (en) | input method and device | |
CN109992124A (en) | Input method, device and machine readable media | |
CN107918496A (en) | It is a kind of to input error correction method and device, a kind of device for being used to input error correction | |
CN108021897A (en) | Picture answering method and device | |
CN108803890A (en) | A kind of input method, input unit and the device for input | |
CN111158487A (en) | Man-machine interaction method for interacting with intelligent terminal by using wireless earphone | |
CN106330864B (en) | The processing method of verification information, apparatus and system | |
CN108182002A (en) | Layout method, device, equipment and the storage medium of enter key | |
CN111639209B (en) | Book content searching method, terminal equipment and storage medium | |
CN107179835A (en) | A kind of input method and device, a kind of device for being used to input | |
CN108073291B (en) | Input method and device and input device | |
CN109977424A (en) | A kind of training method and device of Machine Translation Model | |
CN112488157A (en) | Dialog state tracking method and device, electronic equipment and storage medium |
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 |