CN106534548A - Voice error correction method and device - Google Patents
Voice error correction method and device Download PDFInfo
- Publication number
- CN106534548A CN106534548A CN201611034174.5A CN201611034174A CN106534548A CN 106534548 A CN106534548 A CN 106534548A CN 201611034174 A CN201611034174 A CN 201611034174A CN 106534548 A CN106534548 A CN 106534548A
- Authority
- CN
- China
- Prior art keywords
- error correction
- text data
- word
- content
- error
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M1/00—Substation equipment, e.g. for use by subscribers
- H04M1/72—Mobile telephones; Cordless telephones, i.e. devices for establishing wireless links to base stations without route selection
- H04M1/724—User interfaces specially adapted for cordless or mobile telephones
- H04M1/72403—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality
- H04M1/7243—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality with interactive means for internal management of messages
- H04M1/72436—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality with interactive means for internal management of messages for text messaging, e.g. SMS or e-mail
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M1/00—Substation equipment, e.g. for use by subscribers
- H04M1/72—Mobile telephones; Cordless telephones, i.e. devices for establishing wireless links to base stations without route selection
- H04M1/724—User interfaces specially adapted for cordless or mobile telephones
- H04M1/72403—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality
- H04M1/7243—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality with interactive means for internal management of messages
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M1/00—Substation equipment, e.g. for use by subscribers
- H04M1/72—Mobile telephones; Cordless telephones, i.e. devices for establishing wireless links to base stations without route selection
- H04M1/724—User interfaces specially adapted for cordless or mobile telephones
- H04M1/72403—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality
- H04M1/7243—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality with interactive means for internal management of messages
- H04M1/72439—User interfaces specially adapted for cordless or mobile telephones with means for local support of applications that increase the functionality with interactive means for internal management of messages for image or video messaging
Abstract
The invention provides a voice error correction method and device. The voice error correction method comprises that voice data of a user is received; a preset error correction mode which includes a semantic error correction mode or an index error correction mode is determined; according to the voice data of the user and the present error correction mode, an error of a content to be corrected is corrected; and the content after error correction is back fed to the user. The method can improve the accuracy and suitable range of error correction, requirements of the user can be met more effectively, and user experience is improved.
Description
Technical field
The application is related to natural language understanding technology field, more particularly to a kind of voice error correction method and device.
Background technology
Increasingly mature with artificial intelligence's correlation technique, increasing smart machine enters in the life of access customer, people
It is day by day usual with interacting for machine.Used in interaction, frequency highest is generally interactive voice, and this interactive mode can be with
The both hands of liberation people, so as to be liked by user, such as phonetic entry, voice dialogue.Increasing smart machine is carried now
For the function of voice error correction, user the display content in smart machine can be modified using voice, further liberates
The both hands of people, substantially increase Consumer's Experience.
In correlation technique, the method for voice error correction is usually to carry out error correction to text data, and during concrete error correction, user is necessary
Error correction is carried out according to fixed model, causes to limit more, error correction accuracy is relatively low, it is impossible to meet user's request.
The content of the invention
The application is intended at least to solve to a certain extent one of technical problem in correlation technique.
For this purpose, a purpose of the application is to propose a kind of voice error correction method, it is accurate that the method can improve error correction
Degree and the scope of application, so as to more preferably meet user's request, lift Consumer's Experience.
Further object is to propose a kind of voice error correction device.
For reaching above-mentioned purpose, the voice error correction method that the application first aspect embodiment is proposed, including：Receive user language
Sound data；It is determined that current error correction mode, the error correction mode includes：Semantic error correction mode or index error correction mode；According to institute
User voice data and the current error correction mode is stated, treating error correction content carries out error correction；By the content feed after error correction to use
Family.
The voice error correction method that the application first aspect embodiment is proposed, by determining error correction mode, can select to be adapted to
The error correction mode of current scene, so that improve error correction accuracy；Error correction is carried out by treating error correction content, is not limited to textual data
According to being processed, the scope of application can be extended；Therefore, by improving error correction accuracy and the extension scope of application, more preferably can expire
Sufficient user's request, lifts Consumer's Experience.
For reaching above-mentioned purpose, the voice error correction device that the application second aspect embodiment is proposed, including：Receiver module,
For receive user speech data；Determining module, for determining current error correction mode, the error correction mode includes：Semantic error correction
Pattern or index error correction mode；Correction module, for according to the user voice data and the current error correction mode, treating
Error correction content carries out error correction；Feedback module, for by the content feed after error correction to user.
The voice error correction device that the application second aspect embodiment is proposed, by determining error correction mode, can select to be adapted to
The error correction mode of current scene, so that improve error correction accuracy；Error correction is carried out by treating error correction content, is not limited to textual data
According to being processed, the scope of application can be extended；Therefore, by improving error correction accuracy and the extension scope of application, more preferably can expire
Sufficient user's request, lifts Consumer's Experience.
The aspect and advantage that the application is added will be set forth in part in the description, and partly will become from the following description
Obtain substantially, or recognized by the practice of the application.
Description of the drawings
The above-mentioned and/or additional aspect of the application and advantage will become from the following description of the accompanying drawings of embodiments
It is substantially and easy to understand, wherein：
Fig. 1 is the schematic flow sheet of the voice error correction method that the application one embodiment is proposed；
Fig. 2 is the schematic flow sheet of the voice error correction method that the application another embodiment is proposed；
It is the schematic diagram for treating each word index building in corrected text data in the embodiment of the present application that Fig. 3 is；
Fig. 4 is the schematic diagram for treating corrected text data and corresponding candidate word and candidate index in the embodiment of the present application；
Fig. 5 is the structural representation of the voice error correction device that the application one embodiment is proposed；
Fig. 6 is the structural representation of the voice error correction device that the application another embodiment is proposed；
Fig. 7 is the structural representation of the voice error correction device that the application another embodiment is proposed.
Specific embodiment
Embodiments herein is described below in detail, the example of the embodiment is shown in the drawings, wherein from start to finish
Same or similar label represents same or similar module or the module with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, is only used for explaining the application, and it is not intended that restriction to the application.Conversely, this
The embodiment of application includes all changes, modification and the equivalent fallen in the range of the spirit and intension of attached claims
Thing.
Fig. 1 is the schematic flow sheet of the voice error correction method that the application one embodiment is proposed.
As shown in figure 1, the method for the present embodiment includes：
S11：Receive user speech data.
The user voice data is generally wrong content of the user for showing and carries out the speech data of error correction, described aobvious
The wrong content for showing such as the Error Text data for showing, naturally it is also possible to for other display contents, such as image.
In the present embodiment, in voice error correction, various error correction modes can be divided into, such as be referred to as semantic error correction mode
With index error correction mode, accordingly, speech data when user voice data can be semantic error correction mode, or index
Speech data during error correction mode.
Hypothesis treats that error correction content is text data, and this article notebook data is " Hefei is to Pekinese's train ticket ", and user thinks
What is ordered is Nanjing to Pekinese's train ticket, then user can carry out error correction by voice, when such as user says semantic error correction mode
Speech data, be such as " Nanjing is revised as in Hefei ".Speech data during index error correction mode is index, and index is usually to count
Word is numbered, and such as user says " two points two ".
The user voice data is typically based on user needs the content of modification to determine that particular content is the application do not limit
It is fixed.
S12：It is determined that current error correction mode, the error correction mode includes：Semantic error correction mode or index error correction mode.
In some examples, current error correction mode can be automatically determined by system.
In some examples, the current error correction mode that can be selected by system receive user.
Further, when system automatically determines current error correction mode, can be the history pronunciation or current of system of users
The environment that pronunciation or user are presently in is analyzed, and automatically determines current error correction mode；If user pronunciation is compared with standard, Yong Husuo
When the environment at place is quieter, the quality of user voice data is higher, and semantic understanding accuracy is higher, and system can be automatically determined
Current error correction mode type is semantic error correction mode；If instead user pronunciation is nonstandard or user residing for environment noise compared with
When big, the quality of user voice data is relatively low, and semantic understanding accuracy is not high, and the recognition effect of numeral typically compares Chinese character
Recognition effect is good, then system can automatically determine current error correction mode type for index error correction mode.Or,
When system automatically determines current error correction mode, or the error correction mode that selected according to user's history of system, from
It is dynamic to determine current error correction mode；As user's history generally selects semantic error correction mode, then show user's semanteme accustomed to using error correction
Pattern, then it is semantic error correction mode that system can automatically determine current error correction mode.
During the current error correction mode that system receive user is selected, for example, two kinds of error correction modes alternatively, are passed through by system
The mode such as display or speech play is supplied to user, user to select current error correction mode by the operation such as gesture, voice or button.
It should be noted that, although S11 and S12 are connected with each other in Fig. 1, but this is a kind of example, in actual enforcement
When, current error correction mode can be relevant with the user voice data for receiving, and the user voice data such as to receiving is analyzed,
Judge pronunciation whether standard, whether standard determines current error correction mode according to pronunciation, and now, S11 and S12 can be connected with each other.
Or, current error correction mode can also be unrelated with the user voice data for receiving, and such as, system automatically analyzes user and is presently in
Environment when determining current error correction mode, or when selecting to determine current error correction mode according to user, be not according to receiving
The current error correction mode that user voice data determines, now, S11 and S12 are not connected with each other, and are detached.
S13：According to the user voice data and the current error correction mode, treating error correction content carries out error correction.
Under semantic error correction mode, mainly by carrying out to user voice data after semantic understanding, tied according to semantic understanding
Fruit treats error correction content carries out error correction.Under index error correction mode, mainly index is set up by treating error correction content, user passes through
Treat that the index of error correction content carries out error correction.
It is described to treat that error correction content includes：Text data and non-text data, non-text data are included but is not limited to：Image,
Video, audio frequency, application program.
Under semantic error correction mode, user voice data is usually the corresponding speech data of text data.By taking image as an example,
User voice data is " deleting the 5th image " or " image of one woods will tinkling of pieces of jade of insertion in after second image " etc.；With
As a example by application program, user voice data can be " turning off 360 browsers, open IE browser " etc.；In index error correction mode
Under, after the such as advance video index building to showing, and show the corresponding candidate index of each video, user voice data one
As be the corresponding speech data of candidate index.
S14：By the content feed after error correction to user.
Such as, treat that corrected text data are " Hefei is to Pekinese's train ticket ", user voice data is " Hefei to be revised as
Nanjing ", then, after voice error correction, feed back to user by " Nanjing is to Pekinese's train ticket ".Feedback can be shown by content
Or the mode such as speech play is carried out.
In the present embodiment, by determining error correction mode, can select to be adapted to the error correction mode of current scene, entangle so as to improve
Wrong accuracy；Error correction is carried out by treating error correction content, is not limited to process text data, the scope of application can be extended；
Therefore, by improving error correction accuracy and the extension scope of application, can more preferably meet user's request, lift Consumer's Experience.
Fig. 2 is the schematic flow sheet of the voice error correction method that the application another embodiment is proposed.
The present embodiment is to treat error correction content as treating corrected text data instance.
As shown in Fig. 2 the method for the present embodiment includes：
S21：Receive user speech data.
S22：It is determined that current error correction mode, the error correction mode includes：Semantic error correction mode or index error correction mode.
The particular content of S21-S22 may refer to S11-S12, will not be described in detail herein.
Under different error correction modes, error correction will be carried out using corresponding error correction method.
Specifically, under semantic error correction mode, perform S23-S24；S28 is performed afterwards.Under index error correction mode, perform
S25-S27；S28 is performed afterwards.
S23：Speech recognition is carried out to the user voice data, the corresponding identification textual data of user voice data is obtained
According to.
Speech recognition can be will not be described in detail herein using technology that is various existing or occurring in the future.
S24：Error correction information is determined according to the identification text data, and corrected text number is treated according to the error correction information
According to error correction is carried out, the text data after error correction is obtained.
In some examples, error correction information can be determined according to the identification text data and default error-correction rule, and then
Error correction is carried out using error correction information；The method is properly termed as rule-based method.
In some examples, the identification text data can be extracted and the error correcting characteristics of corrected text data are treated, according to institute
Error correcting characteristics and the voice error correcting model for building in advance is stated, error correction information is determined, and then error correction is carried out using error correction information；The party
Method is properly termed as the method based on model.
The error correction information can include：Erroneous words and error correction term；Error correction term and error correction position；Erroneous words and error correction bit
Put；Or, erroneous words, error correction term and error correction position.
Below above two method is illustrated respectively.
Method one：Rule-based method.
Rule of the rule-based method by predefined voice error correction, directly determines according to the error-correction rule and entangles
Wrong information.The error-correction rule can be predefined according to application demand, and concrete the application is not construed as limiting.
Included as a example by three kinds by error-correction rule, i.e. replaceability error-correction rule, deletion property error-correction rule and insertion property error correction rule
Then, the replaceability error correction needs for erroneous words in text data to be substituted for corresponding error correction term；Insertion property error correction is needed
Error correction term is inserted in corresponding place in text data；Deletion property error correction needs to delete erroneous words in text data.
The following is every kind of error-correction rule example, wherein " * * " represents paraphasia word or error correction term, represent before and after "/" two words be or pass
System：
(1) replaceability error-correction rule：/ make/be modified as * * into * *
Previous " * * " represents erroneous words, and latter " * * " represents error correction term；
(2) insertion property error-correction rule：Behind * */above add/add * *
" behind * */above " error correction position is represented, the position that specifically error correction term is inserted under the rule, latter " * * " table
Show error correction term；
(3) deletion property error-correction rule：Delete/remove behind * */" * * " above
" behind * */above " error correction position is represented, the position that specifically erroneous words are deleted under the rule, latter " * * " table
Show erroneous words.
When error correction information is determined according to identification text data and error-correction rule, can be true according to the identification text data
The error-correction rule being suitable for before settled, and, the identification text data is matched with the current error-correction rule being suitable for,
Determine error correction information.
Specifically, system first judges the current error correction rule being suitable for according to the corresponding identification text data of user voice data
Then, when specifically judging, can be determined according to key word in identification text data, such as " modification " be included in identification text data, " replaced
Change " or during the key word such as " making into ", then can determine that the current error-correction rule being suitable for is replaceability error-correction rule；Then use phase
The error-correction rule (such as replaceability error-correction rule) and identification text data for answering type carries out string matching, determines that error correction is believed
Breath.
According to identification text data and error-correction rule determine error correction information when, it is also possible to it is described identification text data with
Every kind of error-correction rule is matched, and determines error correction information.Directly identification text data is carried out successively with all error-correction rules
The error correction information for matching is defined as the error correction information of final employing by matching.
After determining error correction information, can treat corrected text data according to the error correction information carries out error correction.
During concrete error correction, error correction position can be first determined, when such as error correction information includes erroneous words, by the position of erroneous words
Put as error correction position, or the error correction position that direct access error correction information includes；Respective handling is carried out in error correction position again,
The error correction term such as on error correction position is replaced, or, error correction term is inserted in error correction position, or, in error correction position deletion error word.
It is below rule-based some error correction examples：
(1) replaceability error correction
Treat corrected text data：Hefei is to Pekinese's train ticket
The corresponding identification text data of user voice data：" Hefei " is revised as " Nanjing "
Text data after error correction：Nanjing is to Pekinese's train ticket
(2) insertion property error correction
Treat corrected text data：I wants to play basketball
The corresponding identification text data of user voice data：Eastern school district gymnasium is added before playing basketball
Text data after error correction：I thinks that eastern school district gymnasium is played basketball
(3) deletion property error correction
Treat corrected text data：My phone is May Day 268888
The corresponding identification text data of user voice data：Delete one eight
Text data after error correction：My phone is May Day 26888
Method two：Method based on model.
As default error-correction rule is limited, in order to improve coverage, it is possible to use the error correction side based on model
Method.
In based on the error correction method of model, extract identification text data first and treat the error correcting characteristics of error correction content, then root
According to the feature for extracting and the advance voice error correcting model for building, error correction information is determined.
To treat based on model, error correction content, as corrected text data instance is treated, determines that error correction information can include：
(1) treating corrected text data and identification text data respectively carries out participle.
(2) error correcting characteristics of each word in corrected text data and identification text data are treated described in extracting.
The error correcting characteristics include the position for treating each word in corrected text data, the term vector of each word, each word
The term vector of upper and lower cliction, the mutual information between each word and its context word, the error correction probability of each word, and user speech number
According to the term vector of the term vector of each word, the cliction up and down of each word in identification text data, each word and its context word it
Between mutual information；
Wherein, the cliction up and down of each word refers to the word before each word or word below, specifically considers how many forwards, backwards
Word, can determine according to application demand, such as consider 2 words；Treat that the error correction probability of each word in corrected text data can basis
User's history custom is obtained, and such as user Jing often carries out error correction to a word, then the error correction probability that can arrange the word is larger；Each
Word can be calculated by prior art with the mutual information of its context word.
(3) the voice error correcting model built according to the error correcting characteristics for extracting and in advance, determines error correction information.
During concrete determination, the error correcting characteristics for extracting are output as into phase as the input feature vector of voice error correcting model directly
Answer erroneous words and/or error correction term, and error correction position；For the error correction of insertion property is without output error word, i.e. erroneous words output parameter
For sky, it is only necessary to export error correction term and corresponding error correction position, for the error correction of deletion property is without exporting error correction term, i.e. error correction term
Output parameter is sky, it is only necessary to output error word and corresponding error correction position, for replaceability error correction can export mistake simultaneously
Miss word, error correction term and error correction position.
The voice error correcting model treats corrected text data and corresponding corrected text data in a large number by collection in advance, adopts
The method of deep learning builds and obtains, when specifically building, it is necessary first to which mark treats the error correction position in corrected text data, and treats
Erroneous words and/or error correction term in corrected text data；Then extract and treat each in corrected text data and corrected text data
The error correcting characteristics of word；The error correcting characteristics will be treated into the error correction position of corrected text data as the input of error correcting model finally
And the erroneous words in corrected text and/or error correction term are carried out to error correcting model according to the annotation results as the output of model
Parameter training, the error correcting model such as deep neural network model.
S25：For treating that corrected text data set up candidate word and candidate index.
Can specifically include：
(1) treating corrected text data carries out participle；
Specifically can be realized using technology that is various existing or occurring in the future.
(2) the word index building obtained for participle.
Specifically, directly each word can be numbered according to sequencing, the index as each word will be numbered.
Such as treat that corrected text data are " this are beautiful children's stories ", after participle, can be each word structure
Index as shown in Figure 3 is built, wherein, index of the numeral above each word for corresponding words.
(3) the corresponding word of institute's predicate is determined to word, and determine candidate score of institute's predicate to word；
For example, each word is corresponded to, can be found from dictionary and be there are other words of word to relation with the word, as the word
Word can specifically include to word：The near synonym of the word, homonym and erroneous words etc..
The word of the word pair that the word of one word can be constituted to word according to the word and the word to the candidate score of word is to score meter
Obtain.And the word of each word pair can be calculated to classification according to corresponding word to score, such as, word is to near synonym pair
When, word can be obtained according to the Semantic Similarity Measurement of word centering word to score；Word to for homonym pair when, word can be with to score
Obtained according to the pronunciation Similarity Measure of word centering word；Word to for erroneous words pair when, word can be according to erroneous words to going out to score
Existing frequency is calculated.Specific word can be using technology that is various existing or occurring in the future to the calculation of score.
When word is calculated to the candidate score of word according to word to score, specifically, if what the word and the word were constituted to word
Word to the word pair including multiple classifications, then by the word of the word pair of multiple classifications score is carried out it is cumulative after, using cumulative score as
Candidate score of the word to word.If the word that the word is constituted to word with the word is to the word pair for a classification, by a class
Candidate score of the word of other word pair to score as the word to word.
For example, word is " beautiful ", its near synonym includes " fine " and " beauty ", and " beautiful " and " fine " is constituted
, to being divided into 0.7, the word of the near synonym pair that " beautiful " is constituted with " beauty " is to being divided into 0.5 for the word of near synonym pair；Its homonym
Including " per second ", and the word of homonym pair that " beautiful " and " per second " are constituted is to being divided into 0.6；Its erroneous words include " fine " and
" marvellous ", and the word of erroneous words pair that " beautiful " and " fine " constitute is to being divided into 0.4, the nearly justice that " beautiful " is constituted with " marvellous "
The word of word pair is to being divided into 0.2；According to above-mentioned example, the word that " fine " is constituted with " beautiful " to including near synonym to and erroneous words
Right, i.e., word is multiple to classification, then now, the word by the word of near synonym pair to score with erroneous words pair adds up to score
Afterwards, using cumulative score as " fine " candidate score, i.e. the candidate score of " fine " be 0.7+0.4=1.1；" per second " with
" beautiful " only constitutes homonym pair, then the candidate score of " per second " be the word of the homonym pair of its composition to score, i.e., it is " every
The candidate score of second " is 0.6.
It is understood that each word pair can be found in dictionary in advance, and the word of each word pair is calculated to score,
It is determined that current when corrected text data, directly from calculated word to score in, find and current treat corrected text
Word required for data is to score.Certainly, the mode in line computation is also not precluded from, is such as determined and current is treated corrected text number
According to required word pair, and the word of word pair needed for calculating in real time is to score.
(4) candidate score according to the corresponding word of institute's predicate to word, determines the candidate word of institute's predicate, and according to institute's predicate
Index and the candidate word candidate score, be that the candidate word builds candidate index.
Specifically, can select candidate score higher than predetermined threshold word to word as corresponding words candidate word.
For candidate word index building when, order that can be according to the candidate score of candidate word from big to small is compiled successively
Number, the candidate index as candidate word will be numbered；And when being numbered to candidate word, can be with the corresponding word of the candidate word
Numbering on the basis of carry out.For example, the numbering of a word is 3, then the candidate word of the word is according to candidate score from big to small suitable
Sequence, successively according to 3.1,3.2 ... be numbered.
For example, the candidate word and candidate index for treating the foundation of corrected text data can be as shown in Figure 4.
It is understood that treating that the corresponding candidate word of corrected text data and candidate index can be according to treating corrected text number
According to change and real-time update.
S26：Speech recognition is carried out to the user voice data, the corresponding identification text of the user voice data is obtained
Data, the identification text data include：Error correction is indexed.
Speech recognition can be will not be described in detail herein using technology that is various existing or occurring in the future.
Error correction index refers to the index corresponding to correct word to be used, and such as user says " three points one ".
S27：In error correction content is treated, corresponding alternating content is indexed with the error correction and replace corresponding wrong content, obtained
Text data to after error correction.
When saying " three points one " such as user, as " three points one " corresponding alternating content is " fine ", then replaced with " fine "
Corresponding wrong content is changed, that is, replaces " beautiful ", such that it is able to the content after error correction be obtained for " this is a fine children's stories event
Thing ".
It should be noted that when the error correction term for not having user to need in candidate word, user can be using semantic error correction mould
Formula carries out error correction.
S28：Text data after error correction is fed back to into user.
Such as, treat that corrected text data are " Hefei is to Pekinese's train ticket ", user voice data is " Hefei to be revised as
Nanjing ", then, after voice error correction, feed back to user by " Nanjing is to Pekinese's train ticket ".Feedback can be shown by content
Or the mode such as speech play is carried out.
In the present embodiment, by determining error correction mode, can select to be adapted to the error correction mode of current scene, entangle so as to improve
Wrong accuracy；Error correction is carried out by treating error correction content, is not limited to process text data, the scope of application can be extended；
Therefore, by improving error correction accuracy and the extension scope of application, can more preferably meet user's request, lift Consumer's Experience.Enter one
Step, error correction can also be carried out based on model with rule-based under semantic error correction mode, can further expand and use model
Enclose and accuracy；Under index error correction mode saying index by user carries out error correction, due to only needing user to say numeral, phase
For the mode for saying text, can be user-friendly to, and comparatively numeral is easy to speech recognition, such that it is able to reduce
Implementation complexity.
Fig. 5 is the structural representation of the voice error correction device that the application one embodiment is proposed.
As shown in figure 5, the device 50 of the present embodiment includes：Receiver module 51, determining module 52, correction module 53 and feedback
Module 54.
Receiver module 51, for receive user speech data；
Determining module 52, for determining current error correction mode, the error correction mode includes：Semantic error correction mode or index
Error correction mode；
Correction module 53, for according to the user voice data and the current error correction mode, treating error correction content and entering
Row error correction；
Feedback module 54, for by the content feed after error correction to user.
In some embodiments, referring to Fig. 6, if current error correction mode is semantic error correction mode, the correction module 53 is wrapped
Include：
Speech recognition submodule 5301, for carrying out speech recognition to the user voice data, obtains user's language
The corresponding identification text data of sound data；
Error correction submodule 5302, for determining error correction information according to the identification text data, and believes according to the error correction
Breath treats error correction content carries out error correction, obtains the content after error correction.
In some embodiments, the error correction submodule 5302 for according to it is described identification text data determine error correction information,
Including：
According to the identification text data and default error-correction rule, error correction information is determined；
And/or,
Extract the identification text data and treat the error correcting characteristics of error correction content, according to the error correcting characteristics and advance structure
Voice error correcting model, determine error correction information.
In some embodiments, the error correction submodule 5302 is for according to the identification text data and default error correction rule
Then, determine error correction information, including：
The current error-correction rule being suitable for is determined according to the identification text data, and, to the identification text data with
The current error-correction rule being suitable for is matched, and determines error correction information；Or,
The identification text data is matched with every kind of error-correction rule, error correction information is determined.
In some embodiments, the error correction information includes：
Erroneous words and error correction term；Error correction term and error correction position；Erroneous words and error correction position；Or, erroneous words, error correction term and
Error correction position.
In some embodiments, referring to Fig. 7, if current error correction mode is index error correction mode, the correction module 53 is wrapped
Include：
Setting up submodule 5311, for for treating that error correction content sets up alternating content and candidate index；
Speech recognition submodule 5312, for carrying out speech recognition to the user voice data, obtains user's language
The corresponding identification text data of sound data, the identification text data include：Error correction is indexed；
Error correction submodule 5313, for treating in error correction content, indexes corresponding alternating content with the error correction and replaces right
The wrong content answered, obtains the content after error correction.
In some embodiments, if treating error correction content for treating corrected text data, the alternating content is candidate word, institute
State setting up submodule 5311 specifically for：
Treating corrected text data carries out participle；
For the word index building that participle is obtained；
The corresponding word of institute's predicate is determined to word, and determine candidate score of institute's predicate to word；
According to the index and candidate score of the corresponding word of institute's predicate to word of institute's predicate, the candidate word of institute's predicate is determined, with
And according to the index and the candidate score of the candidate word of institute's predicate, be that the candidate word builds candidate index.
It is in some embodiments, described to treat that error correction content includes：
Text data and non-text data.
It is understood that the device of the present embodiment is corresponding with said method embodiment, particular content may refer to method
The associated description of embodiment, here are no longer described in detail.
In the present embodiment, by determining error correction mode, can select to be adapted to the error correction mode of current scene, entangle so as to improve
Wrong accuracy；Error correction is carried out by treating error correction content, is not limited to process text data, the scope of application can be extended；
Therefore, by improving error correction accuracy and the extension scope of application, can more preferably meet user's request, lift Consumer's Experience.
It is understood that same or similar part mutually can refer in the various embodiments described above, in certain embodiments
Unspecified content may refer to same or analogous content in other embodiment.
It should be noted that in the description of the present application, term " first ", " second " etc. are only used for describing purpose, and not
It is understood that as indicating or implying relative importance.Additionally, in the description of the present application, unless otherwise stated, the implication of " multiple "
Refer at least two.
In flow chart or here any process described otherwise above or method description are construed as, expression includes
It is one or more for realizing specific logical function or process the step of the module of code of executable instruction, fragment or portion
Point, and the scope of the preferred implementation of the application includes other realization, wherein can not be by the suitable of shown or discussion
Sequence, including according to involved function by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be by the application
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each several part of the application can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In embodiment, the software that multiple steps or method can be performed in memory and by suitable instruction execution system with storage
Or firmware is realizing.For example, if realized with hardware, and in another embodiment, can be with well known in the art
Any one of row technology or their combination are realizing：With for realizing the logic gates of logic function to data signal
Discrete logic, the special IC with suitable combinational logic gate circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method is carried
Suddenly the hardware that can be by program to instruct correlation is completed, and described program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
Additionally, each functional unit in the application each embodiment can be integrated in a processing module, it is also possible to
It is that unit is individually physically present, it is also possible to which two or more units are integrated in a module.Above-mentioned integrated mould
Block both can be realized in the form of hardware, it would however also be possible to employ the form of software function module is realized.The integrated module is such as
Fruit using in the form of software function module realize and as independent production marketing or use when, it is also possible to be stored in a computer
In read/write memory medium.
Storage medium mentioned above can be read only memory, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
Example ", or the description of " some examples " etc. mean specific features with reference to the embodiment or example description, structure, material or spy
Point is contained at least one embodiment or example of the application.In this manual, to the schematic representation of above-mentioned term not
Identical embodiment or example are referred to necessarily.And, the specific features of description, structure, material or feature can be any
One or more embodiments or example in combine in an appropriate manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is impossible to the restriction to the application is interpreted as, one of ordinary skill in the art within the scope of application can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (16)
1. a kind of voice error correction method, it is characterised in that include：
Receive user speech data；
It is determined that current error correction mode, the error correction mode includes：Semantic error correction mode or index error correction mode；
According to the user voice data and the current error correction mode, treating error correction content carries out error correction；
By the content feed after error correction to user.
2. method according to claim 1, it is characterised in that if current error correction mode is semantic error correction mode, described
According to the user voice data and the current error correction mode, treating error correction content carries out error correction, including：
Speech recognition is carried out to the user voice data, the corresponding identification text data of the user voice data is obtained；
Determining error correction information according to the identification text data, and treat error correction content according to the error correction information carries out error correction,
Obtain the content after error correction.
3. method according to claim 2, it is characterised in that described to determine that error correction is believed according to the identification text data
Breath, including：
According to the identification text data and default error-correction rule, error correction information is determined；
And/or,
Extract the identification text data and treat the error correcting characteristics of error correction content, according to the error correcting characteristics and the advance language for building
Sound error correcting model, determines error correction information.
4. method according to claim 3, it is characterised in that described according to the identification text data and default error correction
Rule, determines error correction information, including：
The current error-correction rule being suitable for is determined according to the identification text data, and, to the identification text data with it is described
The current error-correction rule being suitable for is matched, and determines error correction information；Or,
The identification text data is matched with every kind of error-correction rule, error correction information is determined.
5. the method according to any one of claim 2-4, it is characterised in that the error correction information includes：
Erroneous words and error correction term；Error correction term and error correction position；Erroneous words and error correction position；Or, erroneous words, error correction term and error correction
Position.
6. method according to claim 1, it is characterised in that if current error correction mode is index error correction mode, described
According to the user voice data and the current error correction mode, treating error correction content carries out error correction, including：
For treating that error correction content sets up alternating content and candidate index；
Speech recognition is carried out to the user voice data, the corresponding identification text data of the user voice data, institute is obtained
Stating identification text data includes：Error correction is indexed；
In error correction content is treated, corresponding alternating content is indexed with the error correction and replace corresponding wrong content, after obtaining error correction
Content.
7. method according to claim 6, it is characterised in that if treating error correction content for treating corrected text data, institute
It is candidate word to state alternating content, described for treating that error correction content sets up alternating content and candidate index, including：
Treating corrected text data carries out participle；
For the word index building that participle is obtained；
The corresponding word of institute's predicate is determined to word, and determine candidate score of institute's predicate to word；
According to candidate score of the corresponding word of institute's predicate to word, the candidate word of institute's predicate is determined, and according to the index of institute's predicate
With the candidate score of the candidate word, it is that the candidate word builds candidate index.
8. method according to claim 1, it is characterised in that described to treat that error correction content includes：
Text data and non-text data.
9. a kind of voice error correction device, it is characterised in that include：
Receiver module, for receive user speech data；
Determining module, for determining current error correction mode, the error correction mode includes：Semantic error correction mode or index error correction mould
Formula；
Correction module, for according to the user voice data and the current error correction mode, treating error correction content carries out error correction；
Feedback module, for by the content feed after error correction to user.
10. device according to claim 9, it is characterised in that if current error correction mode is semantic error correction mode, described
Correction module includes：
Speech recognition submodule, for carrying out speech recognition to the user voice data, obtains the user voice data pair
The identification text data answered；
Error correction submodule, for determining error correction information according to the identification text data, and treats according to the error correction information and entangles
Wrong content carries out error correction, obtains the content after error correction.
11. devices according to claim 10, it is characterised in that the error correction submodule is for according to the identification text
Data determine error correction information, including：
According to the identification text data and default error-correction rule, error correction information is determined；
And/or,
Extract the identification text data and treat the error correcting characteristics of error correction content, according to the error correcting characteristics and the advance language for building
Sound error correcting model, determines error correction information.
12. devices according to claim 10, it is characterised in that the error correction submodule is for according to the identification text
Data and default error-correction rule, determine error correction information, including：
The current error-correction rule being suitable for is determined according to the identification text data, and, to the identification text data with it is described
The current error-correction rule being suitable for is matched, and determines error correction information；Or,
The identification text data is matched with every kind of error-correction rule, error correction information is determined.
13. methods according to any one of claim 10-12, it is characterised in that the error correction information includes：
Erroneous words and error correction term；Error correction term and error correction position；Erroneous words and error correction position；Or, erroneous words, error correction term and error correction
Position.
14. methods according to claim 9, it is characterised in that if current error correction mode is index error correction mode, described
Correction module includes：
Setting up submodule, for for treating that error correction content sets up alternating content and candidate index；
Speech recognition submodule, for carrying out speech recognition to the user voice data, obtains the user voice data pair
The identification text data answered, the identification text data include：Error correction is indexed；
Error correction submodule, for treating in error correction content, indexes corresponding alternating content with the error correction and replaces corresponding mistake
Content, obtains the content after error correction.
15. devices according to claim 14, it is characterised in that if treating error correction content for treating corrected text data,
The alternating content be candidate word, the setting up submodule specifically for：
Treating corrected text data carries out participle；
For the word index building that participle is obtained；
The corresponding word of institute's predicate is determined to word, and determine candidate score of institute's predicate to word；
According to candidate score of the corresponding word of institute's predicate to word, the candidate word of institute's predicate is determined, and according to the index of institute's predicate
With the candidate score of the candidate word, it is that the candidate word builds candidate index.
16. devices according to claim 9, it is characterised in that described to treat that error correction content includes：
Text data and non-text data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611034174.5A CN106534548B (en) | 2016-11-17 | 2016-11-17 | Voice error correction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611034174.5A CN106534548B (en) | 2016-11-17 | 2016-11-17 | Voice error correction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106534548A true CN106534548A (en) | 2017-03-22 |
| CN106534548B CN106534548B (en) | 2020-06-12 |
Family
ID=58357791
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611034174.5A Active CN106534548B (en) | 2016-11-17 | 2016-11-17 | Voice error correction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106534548B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106992001A (en) * | 2017-03-29 | 2017-07-28 | 百度在线网络技术（北京）有限公司 | Processing method, the device and system of phonetic order |
| CN107221328A (en) * | 2017-05-25 | 2017-09-29 | 百度在线网络技术（北京）有限公司 | The localization method and device in modification source, computer equipment and computer-readable recording medium |
| CN107463601A (en) * | 2017-06-13 | 2017-12-12 | 北京百度网讯科技有限公司 | Dialogue based on artificial intelligence understands system constituting method, device, equipment and computer-readable recording medium |
| CN107480118A (en) * | 2017-08-16 | 2017-12-15 | 科大讯飞股份有限公司 | Method for editing text and device |
| CN107544726A (en) * | 2017-07-04 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | Method for correcting error of voice identification result, device and storage medium based on artificial intelligence |
| CN107832447A (en) * | 2017-11-22 | 2018-03-23 | 北京百度网讯科技有限公司 | User feedback error correction method, device and its equipment for mobile terminal |
| CN109215660A (en) * | 2018-07-09 | 2019-01-15 | 维沃移动通信有限公司 | Text error correction method and mobile terminal after speech recognition |
| CN109243433A (en) * | 2018-11-06 | 2019-01-18 | 北京百度网讯科技有限公司 | Audio recognition method and device |
| CN109389977A (en) * | 2018-11-01 | 2019-02-26 | 腾讯大地通途（北京）科技有限公司 | A kind of voice interactive method and device |
| CN109994105A (en) * | 2017-12-29 | 2019-07-09 | 宝马股份公司 | Data inputting method, device, system, vehicle and readable storage medium storing program for executing |
| CN110211577A (en) * | 2019-07-19 | 2019-09-06 | 宁波方太厨具有限公司 | Terminal device and its voice interactive method |
| CN110413445A (en) * | 2018-04-28 | 2019-11-05 | 北京搜狗科技发展有限公司 | Input processing method, device, electronic equipment and storage medium |
| CN110415679A (en) * | 2019-07-25 | 2019-11-05 | 北京百度网讯科技有限公司 | Voice error correction method, device, equipment and storage medium |
| CN110717021A (en) * | 2019-09-17 | 2020-01-21 | 平安科技（深圳）有限公司 | Input text and related device for obtaining artificial intelligence interview |
| CN110781665A (en) * | 2019-10-29 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Method, device and equipment for evaluating quality of error correction pair and storage medium |
| CN110929875A (en) * | 2019-10-12 | 2020-03-27 | 平安国际智慧城市科技股份有限公司 | Intelligent language learning method, system, device and medium based on machine learning |
| CN111626049A (en) * | 2020-05-27 | 2020-09-04 | 腾讯科技（深圳）有限公司 | Title correction method and device for multimedia information, electronic equipment and storage medium |
| CN112365892A (en) * | 2020-11-10 | 2021-02-12 | 杭州大搜车汽车服务有限公司 | Man-machine interaction method, device, electronic device and storage medium |
| CN116229975A (en) * | 2023-03-17 | 2023-06-06 | 杭州盈禾嘉田科技有限公司 | System and method for voice reporting of field diseases and insect pests in intelligent interaction scene |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1979638A (en) * | 2005-12-02 | 2007-06-13 | 中国科学院自动化研究所 | Method for correcting error of voice identification result |
| CN101655837A (en) * | 2009-09-08 | 2010-02-24 | 北京邮电大学 | Method for detecting and correcting error on text after voice recognition |
| CN104464736A (en) * | 2014-12-15 | 2015-03-25 | 北京百度网讯科技有限公司 | Error correction method and device for voice recognition text |
| CN104978964A (en) * | 2014-04-14 | 2015-10-14 | 美的集团股份有限公司 | Voice control instruction error correction method and system |
-
2016
- 2016-11-17 CN CN201611034174.5A patent/CN106534548B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1979638A (en) * | 2005-12-02 | 2007-06-13 | 中国科学院自动化研究所 | Method for correcting error of voice identification result |
| CN101655837A (en) * | 2009-09-08 | 2010-02-24 | 北京邮电大学 | Method for detecting and correcting error on text after voice recognition |
| CN104978964A (en) * | 2014-04-14 | 2015-10-14 | 美的集团股份有限公司 | Voice control instruction error correction method and system |
| CN104464736A (en) * | 2014-12-15 | 2015-03-25 | 北京百度网讯科技有限公司 | Error correction method and device for voice recognition text |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106992001B (en) * | 2017-03-29 | 2020-05-22 | 百度在线网络技术（北京）有限公司 | Voice instruction processing method, device and system |
| CN106992001A (en) * | 2017-03-29 | 2017-07-28 | 百度在线网络技术（北京）有限公司 | Processing method, the device and system of phonetic order |
| CN107221328B (en) * | 2017-05-25 | 2021-02-19 | 百度在线网络技术（北京）有限公司 | Method and device for positioning modification source, computer equipment and readable medium |
| CN107221328A (en) * | 2017-05-25 | 2017-09-29 | 百度在线网络技术（北京）有限公司 | The localization method and device in modification source, computer equipment and computer-readable recording medium |
| US10528670B2 (en) | 2017-05-25 | 2020-01-07 | Baidu Online Network Technology (Beijing) Co., Ltd. | Amendment source-positioning method and apparatus, computer device and readable medium |
| CN107463601A (en) * | 2017-06-13 | 2017-12-12 | 北京百度网讯科技有限公司 | Dialogue based on artificial intelligence understands system constituting method, device, equipment and computer-readable recording medium |
| CN107463601B (en) * | 2017-06-13 | 2021-02-12 | 北京百度网讯科技有限公司 | Dialog understanding system construction method, device and equipment based on artificial intelligence and computer readable storage medium |
| US11727302B2 (en) | 2017-06-13 | 2023-08-15 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and apparatus for building a conversation understanding system based on artificial intelligence, device and computer-readable storage medium |
| CN107544726A (en) * | 2017-07-04 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | Method for correcting error of voice identification result, device and storage medium based on artificial intelligence |
| CN107480118A (en) * | 2017-08-16 | 2017-12-15 | 科大讯飞股份有限公司 | Method for editing text and device |
| CN107832447A (en) * | 2017-11-22 | 2018-03-23 | 北京百度网讯科技有限公司 | User feedback error correction method, device and its equipment for mobile terminal |
| CN109994105A (en) * | 2017-12-29 | 2019-07-09 | 宝马股份公司 | Data inputting method, device, system, vehicle and readable storage medium storing program for executing |
| CN110413445A (en) * | 2018-04-28 | 2019-11-05 | 北京搜狗科技发展有限公司 | Input processing method, device, electronic equipment and storage medium |
| CN110413445B (en) * | 2018-04-28 | 2024-02-02 | 北京搜狗科技发展有限公司 | Input processing method, input processing device, electronic equipment and storage medium |
| CN109215660A (en) * | 2018-07-09 | 2019-01-15 | 维沃移动通信有限公司 | Text error correction method and mobile terminal after speech recognition |
| CN109389977B (en) * | 2018-11-01 | 2021-07-16 | 腾讯大地通途（北京）科技有限公司 | Voice interaction method and device |
| CN109389977A (en) * | 2018-11-01 | 2019-02-26 | 腾讯大地通途（北京）科技有限公司 | A kind of voice interactive method and device |
| CN109243433A (en) * | 2018-11-06 | 2019-01-18 | 北京百度网讯科技有限公司 | Audio recognition method and device |
| CN109243433B (en) * | 2018-11-06 | 2021-07-09 | 北京百度网讯科技有限公司 | Speech recognition method and device |
| CN110211577A (en) * | 2019-07-19 | 2019-09-06 | 宁波方太厨具有限公司 | Terminal device and its voice interactive method |
| CN110211577B (en) * | 2019-07-19 | 2021-06-04 | 宁波方太厨具有限公司 | Terminal equipment and voice interaction method thereof |
| US11328708B2 (en) | 2019-07-25 | 2022-05-10 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Speech error-correction method, device and storage medium |
| CN110415679B (en) * | 2019-07-25 | 2021-12-17 | 北京百度网讯科技有限公司 | Voice error correction method, device, equipment and storage medium |
| CN110415679A (en) * | 2019-07-25 | 2019-11-05 | 北京百度网讯科技有限公司 | Voice error correction method, device, equipment and storage medium |
| CN110717021A (en) * | 2019-09-17 | 2020-01-21 | 平安科技（深圳）有限公司 | Input text and related device for obtaining artificial intelligence interview |
| CN110717021B (en) * | 2019-09-17 | 2023-08-29 | 平安科技（深圳）有限公司 | Input text acquisition and related device in artificial intelligence interview |
| CN110929875A (en) * | 2019-10-12 | 2020-03-27 | 平安国际智慧城市科技股份有限公司 | Intelligent language learning method, system, device and medium based on machine learning |
| CN110781665B (en) * | 2019-10-29 | 2023-04-07 | 腾讯科技（深圳）有限公司 | Method, device and equipment for evaluating quality of error correction pair and storage medium |
| CN110781665A (en) * | 2019-10-29 | 2020-02-11 | 腾讯科技（深圳）有限公司 | Method, device and equipment for evaluating quality of error correction pair and storage medium |
| CN111626049B (en) * | 2020-05-27 | 2022-12-16 | 深圳市雅阅科技有限公司 | Title correction method and device for multimedia information, electronic equipment and storage medium |
| CN111626049A (en) * | 2020-05-27 | 2020-09-04 | 腾讯科技（深圳）有限公司 | Title correction method and device for multimedia information, electronic equipment and storage medium |
| CN112365892A (en) * | 2020-11-10 | 2021-02-12 | 杭州大搜车汽车服务有限公司 | Man-machine interaction method, device, electronic device and storage medium |
| CN116229975A (en) * | 2023-03-17 | 2023-06-06 | 杭州盈禾嘉田科技有限公司 | System and method for voice reporting of field diseases and insect pests in intelligent interaction scene |
| CN116229975B (en) * | 2023-03-17 | 2023-08-18 | 杭州盈禾嘉田科技有限公司 | System and method for voice reporting of field diseases and insect pests in intelligent interaction scene |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106534548B (en) | 2020-06-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106534548A (en) | Voice error correction method and device | |
| CN109840287B (en) | Cross-modal information retrieval method and device based on neural network | |
| CN110457688B (en) | Error correction processing method and device, storage medium and processor | |
| CN105869634B (en) | It is a kind of based on field band feedback speech recognition after text error correction method and system | |
| WO2021203581A1 (en) | Key information extraction method based on fine annotation text, and apparatus and storage medium | |
| CN107797984B (en) | Intelligent interaction method, equipment and storage medium | |
| US9190054B1 (en) | Natural language refinement of voice and text entry | |
| TWI437449B (en) | Multi-mode input method and input method editor system | |
| CN107077843A (en) | Session control and dialog control method | |
| CN111833853B (en) | Voice processing method and device, electronic equipment and computer readable storage medium | |
| US20140358539A1 (en) | Method and apparatus for building a language model | |
| CN107301860A (en) | Audio recognition method and device based on Chinese and English mixing dictionary | |
| CN108711420A (en) | Multilingual hybrid model foundation, data capture method and device, electronic equipment | |
| CN109637537B (en) | Method for automatically acquiring annotated data to optimize user-defined awakening model | |
| CN107195295A (en) | Audio recognition method and device based on Chinese and English mixing dictionary | |
| CN106886516A (en) | The method and device of automatic identification statement relationship and entity | |
| WO2014190732A1 (en) | Method and apparatus for building a language model | |
| WO2021179701A1 (en) | Multilingual speech recognition method and apparatus, and electronic device | |
| CN109165380A (en) | A kind of neural network model training method and device, text label determine method and device | |
| CN108509409A (en) | A method of automatically generating semantic similarity sentence sample | |
| CN107369439A (en) | A kind of voice awakening method and device | |
| CN112487139A (en) | Text-based automatic question setting method and device and computer equipment | |
| CN106649253A (en) | Auxiliary control method and system based on post verification | |
| CN109543176A (en) | A kind of abundant short text semantic method and device based on figure vector characterization | |
| CN113239666A (en) | Text similarity calculation method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |