CN108847241A - It is method, electronic equipment and the storage medium of text by meeting speech recognition - Google Patents
It is method, electronic equipment and the storage medium of text by meeting speech recognition Download PDFInfo
- Publication number
- CN108847241A CN108847241A CN201810581922.4A CN201810581922A CN108847241A CN 108847241 A CN108847241 A CN 108847241A CN 201810581922 A CN201810581922 A CN 201810581922A CN 108847241 A CN108847241 A CN 108847241A
- Authority
- CN
- China
- Prior art keywords
- text
- word
- speech recognition
- default
- database
- 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
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000005516 engineering process Methods 0.000 claims abstract description 21
- 239000002245 particle Substances 0.000 claims description 53
- 230000004048 modification Effects 0.000 claims description 39
- 238000012986 modification Methods 0.000 claims description 39
- 238000004590 computer program Methods 0.000 claims description 18
- 238000013135 deep learning Methods 0.000 claims description 18
- 238000012790 confirmation Methods 0.000 claims description 17
- 206010028916 Neologism Diseases 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 12
- 230000000717 retained effect Effects 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 3
- 230000001915 proofreading effect Effects 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 15
- 230000008569 process Effects 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000006399 behavior Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 208000001613 Gambling Diseases 0.000 description 2
- 230000002411 adverse Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 150000001875 compounds Chemical class 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Machine Translation (AREA)
- Telephonic Communication Services (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
- Document Processing Apparatus (AREA)
Abstract
A method of by meeting speech recognition be text, including：Text is converted by conference voice to be identified by speech recognition technology, identifies text as initial speech；Initial speech identification text is matched with pre-set text database, the speech recognition text after being matched；There is the speech recognition text rough draft of editable state according to the speech recognition text generation after matching；When detect have received edit operation on speech recognition text rough draft after, according to the speech recognition text generation after the edit operation have can not editing mode speech recognition text, as final speech recognition text.It is the electronic equipment and storage medium of text that the present invention also provides a kind of by meeting speech recognition.The present invention carries out first time matching by carrying out after tentatively identifying to voice to be identified, with pre-set text library, then is confirmed by manually carrying out second, and the effective correctness for guaranteeing text output content reduces the proof-reading amount of conference content, improves efficiency.
Description
Technical field
The present invention relates to technical field of voice recognition, and in particular to a kind of by meeting speech recognition is the method for text, electricity
Sub- equipment and storage medium.
Background technique
Automatic speech recognition technology (Automatic Speech Recognition, ASR) is machine translation, robot control
Computer " can be dictated " and goes out different people and is said for the core technology in the fields such as system and next-generation human-computer interaction interface
Continuous speech realizes that " sound " arrives the conversion of " text ".
Currently, the application based on speech recognition is also more and more extensive, in this way along with the continuous development of speech recognition technology
Technology infiltrated through family life, office realm, amusement etc..User, which passes through, utilizes opposite personal computer, notes
External or built-in microphone inputs voice in the terminals such as this computer, tablet computer, dedicated learning terminal, smart phone,
The conversion of voice-text is completed via speech recognition apparatus.
Existing speech recognition apparatus has very much, for example, the language of world-famous Nuance, Google for being widely used
Sound identification service, China hold speech-recognition services of Iflytek etc..But it is in the greatest problem for carrying out speech recognition
The accuracy rate of speech recognition can not also be kept away even possessing the Nuance of highest speech recognition accuracy in existing equipment
Exempt from following problems：Frequently occurring for the unrelated vocabulary such as modal particle causes text analyzing difficulty to increase, the professional keyword identification in part
Inaccuracy, taboo sensitive word can not identify, affect the readable analytical of meeting text.
Summary of the invention
In view of the foregoing, it is necessary to propose a kind of method, electronic equipment and storage by meeting speech recognition for text
Text output content is being effectively guaranteed just by the double process of pre-set text database matching and manual confirmation in medium
True property improves traditional voice and turns the unreasonable place of literal expression in text, effectively reduces the proof-reading of conference content
Amount, improves efficiency.
The first aspect of the present invention provides a kind of method by meeting speech recognition for text, the method includes：
Text is converted by conference voice to be identified by speech recognition technology, identifies text as initial speech；
Initial speech identification text is matched with pre-set text database, the speech recognition text after being matched
This；
There is the speech recognition text rough draft of editable state according to the speech recognition text generation after the matching；
When detect have received edit operation on the speech recognition text rough draft when, after the edit operation
Speech recognition text generation have can not editing mode speech recognition text, as final speech recognition text.
Preferably, it is described by the initial speech identification text and pre-set text database match including：
Initial speech identification text is matched with default first text database, obtains the first matching result；
First matching result is matched with default second text database, obtains the second matching result；
Second matching result is matched with default third text database；
Wherein, multiple modal particles, default second text database are stored in default first text database
In be stored with multiple professional words and corresponding phonetic, be stored with multiple taboo sensitive words in the default third text database.
Preferably, it is described by the initial speech identification text and default first text database match including：
Judge to match in the initial speech identification text with the presence or absence of with the word in default first text database
The first word；
When determine exist in initial speech identification text and default first text database in word match
When the first word, according to the first word for matching described in the modal particle model judgement based on deep learning network of training in advance
It whether is modal particle to be deleted；
When the first word to match described in the determination is modal particle to be deleted, the initial speech is identified in text
The first word to match is rejected；
When the first word to match described in the determination is not modal particle to be deleted, the initial speech is identified into text
In the first word for matching retained.
Preferably, it is described by first matching result and default second text database match including：
Word in first matching result is converted into the first phonetic；
Judge in default second text database with the presence or absence of the second phonetic identical with first phonetic；
When determine there is the second phonetic identical with first phonetic in default second text database when, by the
The corresponding word of two phonetics extracts, as the corresponding word of the first phonetic.
Preferably, it is described by second matching result and default third text database match including：
Judge in second matching result with the presence or absence of the to match with the word in default third text database
Three words；
There is the third that matches with the word in default third text database in second matching result when determining
When word, the third word to match in second matching result is rejected.
Preferably, the speech recognition text generation according to after the edit operation have can not editing mode voice
Identify that text includes：
When the edit operation received be confirmation operation when, directly generate with can not editing mode speech recognition text
This；
When the edit operation received is modification operation, receives the manual modification of user and saves the new content of modification,
When receiving confirmation operation again, generate have can not editing mode speech recognition text.
Preferably, the method also includes：
By at corresponding modification original word and the modified neologism of user be associated storage；
When subsequently through speech recognition technology, text is converted for conference voice to be identified according to modified neologism
This.
Preferably, the method also includes：
The corresponding diversified forms of each word are stored in advance, the diversified forms include：Simplified and traditional body form plus space form
And nearly word form；
By the diversified forms progress corresponding with each word of pre-set text database of initial speech identification text
Match.
The second aspect of the present invention provides a kind of electronic equipment, and the electronic equipment includes processor and memory, described
Processor is used to realize the side by meeting speech recognition for text in the memory when executing the computer program stored
Method.
The third aspect of the present invention provides a kind of computer readable storage medium, deposits on the computer readable storage medium
Computer program is contained, the computer program realizes the side by meeting speech recognition for text when being executed by processor
Method.
It is of the present invention by meeting speech recognition be text method, electronic equipment and storage medium, known by voice
Conference voice to be identified is converted text by other technology, identifies text as initial speech；The initial speech is identified into text
This is matched with pre-set text database, the speech recognition text after being matched；According to the speech recognition after the matching
Text generation has the speech recognition text rough draft of editable state；It is received on the speech recognition text rough draft when detecting
When edit operation, according to speech recognition text generation after the edit operation have can not editing mode speech recognition text
This, as final speech recognition text.The present invention by voice to be identified carry out tentatively identify after, with pre-set text library into
Row matches for the first time, then is confirmed by manually carrying out second.Process can effectively ensure that text output content just twice
True property improves traditional voice and turns the unreasonable place of literal expression in text, effectively reduces the proof-reading of conference content
Amount, improves efficiency.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, 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
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the flow chart by the method that meeting speech recognition is text that the embodiment of the present invention one provides.
Fig. 2 is the functional block diagram provided by Embodiment 2 of the present invention by the device that meeting speech recognition is text.
Fig. 3 is the schematic diagram for the electronic equipment that the embodiment of the present invention three provides.
The present invention that the following detailed description will be further explained with reference to the above drawings.
Specific embodiment
To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real
Applying example, the present invention will be described in detail.It should be noted that in the absence of conflict, the embodiment of the present invention and embodiment
In feature can be combined with each other.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, described embodiment is only
It is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill
Personnel's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Unless otherwise defined, all technical and scientific terms used herein and belong to technical field of the invention
The normally understood meaning of technical staff is identical.Term as used herein in the specification of the present invention is intended merely to description tool
The purpose of the embodiment of body, it is not intended that in the limitation present invention.
The embodiment of the present invention applies the method that meeting speech recognition is text in one or more electronic equipment.
It is described that the method that meeting speech recognition is text also can be applied to by electronic equipment and pass through network and the electronic equipment
In the hardware environment that the server being attached is constituted.Network includes but is not limited to：Wide area network, Metropolitan Area Network (MAN) or local area network.This
Can the executing the method that meeting speech recognition is text by server of inventive embodiments, can also be held by electronic equipment
Row；It can also be and executed jointly by server and electronic equipment.
It is described that needs can directly set the electronic equipment for the method that meeting speech recognition is text in electronics
The function provided by method of the invention by meeting speech recognition for text, or installation are integrated on standby for realizing the present invention
Method client.For another example, method provided by the present invention can also be with Software Development Kit (Software
Development Kit, SDK) form operate in the equipment such as server, provided in the form of SDK by meeting speech recognition
For the interface of the function of text, it is text that the interface that electronic equipment or other equipment pass through offer, which can be realized meeting speech recognition,
This.
Embodiment one
Fig. 1 is the flow chart by the method that meeting speech recognition is text that the embodiment of the present invention one provides.According to difference
Demand, the execution sequence in the flow chart can change, and certain steps can be omitted.
S11, text is converted for conference voice to be identified by speech recognition technology, identifies text as initial speech
This.
In the present embodiment, the detailed process for converting conference voice to be identified to by speech recognition technology in text
Including：
1) audio frequency characteristics for extracting conference voice to be identified, are converted to the acoustic feature vector of preset length；
2) described eigenvector is decoded by word order according to decoding algorithm；
3) the sub- word of corresponding word order is obtained by HMM phoneme model, the sub- word is initial consonant and simple or compound vowel of a Chinese syllable；
4) multiple sub- words are spliced by text according to preset Pronounceable dictionary；
5) it decodes to obtain optimal sequence using language model syntax rule, obtains text.
The syntax rule is viterbi algorithm.For example, the conference voice to be identified is " hello ", by spy
Sign is converted into the acoustic feature vectors of 39 dimensions after extracting, by multiple HMM phoneme models obtain corresponding multiple sub- word/n//i//
Multiple sub- words are spliced into word according to preset Pronounceable dictionary by h//ao/, as you, Buddhist nun；It is good, number.It is decoded by viterbi algorithm
Obtain optimal sequence " hello " and by text output.
S12, initial speech identification text is matched with pre-set text database, the voice after being matched is known
Other text.
In the present embodiment, at least three text databases can be preset, for example, the first text database, the second text
Database and third circumferential edge library.First text database can be exclusively used in storing multiple modal particles, such as " uh ", " ",
" right " etc., modal particle is unrelated with conference content, and easily influences voice again and be converted to the readability after text.Second text data
Library can be exclusively used in storing multiple professional words and corresponding phonetic, such as " feature vector ", " eigenmatrix ", " tensor analysis ",
Professional word is more complex, thus is easy batch during identifying voice and mistake occur.Third text database can be exclusively used in
Multiple taboo sensitive words are stored, such as politics relevant name, porns, gambling and drugs, firearms and ammunition class, swears at people that it is non-to satirize class button at supertition heresy
The words such as method info class, the appearance for avoiding sensitive word be easy to cause adverse effect.The present invention can also be set in advance according to the actual situation
The 4th text database etc. is set, the sentence of storage name or place name etc. is exclusively used in.Herein for pre-set this paper
The quantity of database and corresponding content are not specifically limited.
It is described match specifically including with pre-set text database by initial speech identification text：
1) initial speech identification text is matched with default first text database, obtains the first matching knot
Fruit；
2) first matching result is matched with default second text database, obtains the second matching result；
3) second matching result is matched with default third text database.
Three independently operated threads can be preset：First thread, the second thread and third thread, first thread are held
It is about to the initial speech identification text to refer to the program that default first text database is matched to obtain the first matching result
It enables, first matching result is matched to obtain the second matching result by the execution of the second thread with default second text database
Program instruction, third thread, which executes, refers to second matching result and the matched program of default third text database progress
It enables.After first thread is finished, the second thread is executed at once, after the second thread is finished, executes third line at once
Journey, three independently operated threads of setting execute different program instructions respectively, are conducive to improve matched speed, save matching
Time.
In other embodiments, a thread can also be only set, and being sequentially carried out by one thread will be described initial
Speech recognition text is matched to obtain with default first text database the program instruction of the first matching result, by described first
Matching result is matched to obtain the program instruction of the second matching result, be matched described second with default second text database
As a result matched program instruction is carried out with default third text database.
Specifically, it is described by the initial speech identification text and default first text database match including：Sentence
The initial speech of breaking identifies in text with the presence or absence of the first word to match with the word in default first text database；
When determining in initial speech identification text there is the first word to match with the word in default first text database
When, the initial speech is identified that the first word to match in text is handled.
Preferably, it is described the initial speech is identified in text that the first word for matching carries out processing can also be into one
Step includes：The first word to match described in modal particle model judgement according to training in advance based on deep learning network
It whether is modal particle to be deleted；When the first word to match described in the determination is modal particle to be deleted, will it is described initially
The first word to match in speech recognition text is rejected；The first word to match described in the determination is not to be deleted
When modal particle, the initial speech is identified that the first word to match in text retains.
For example, it is assumed that initial speech identifies that text is " this is pretty good ", deposits in default first text database
Modal particle " this " is contained, then phase has been determined after being matched initial speech identification text with default first text database
The word matched is " this ", then judges the phase according to the modal particle model based on deep learning network of training in advance
Whether matched first word " this " is modal particle to be deleted, and the modal particle model based on deep learning network determines
The first word " this " to match is not belonging to modal particle to be deleted in " this is pretty good ", then by the initial speech
The first word to match in identification text is retained, and the first obtained matching result is " this is pretty good ".
For another example, it is assumed that initial speech identifies that text is " this, we will have a meeting ", deposits in default first text database
Modal particle " this " is contained, then phase has been determined after being matched initial speech identification text with default first text database
The word matched is " this ", then judges the phase according to the modal particle model based on deep learning network of training in advance
Whether matched first word " this " is modal particle to be deleted, and the modal particle model based on deep learning network determines
The first word " this " to match belongs to modal particle to be deleted in " this, we will have a meeting ", then will be described initial
The first word to match in speech recognition text is rejected, and the first obtained matching result is " we will have a meeting ".
Preferably, the training method of the modal particle model based on deep learning network may include：
1) text largely with the word in the first text database is obtained；
2) text is divided into positive sample text and negative sample text, the positive sample text is to need to retain the tone
The text of word, the negative sample text are the text for needing to delete modal particle；
For example, the word in the first text database is " this ", then it is available largely to contain " this " word
This paper of language, such as " this project is currently ongoing ", " whom this people is ", " this, uh, also in inquiry ", " this can
With such ", wherein " this " is the language for needing to retain in " this project is currently ongoing " and " whom this people is "
Gas word, " this " in " this, uh, also in inquiry ", " this can be such " is the modal particle for needing to delete.
3) the positive sample text is stamped into first identifier, the negative sample text is stamped into second identifier；
The word that the first identifier is used to identify in sample needs to retain, for example, it may be " 1 ".The second identifier
Word for identifying in sample needs to delete, for example, it may be " 0 ".
4) the positive sample text input is trained into deep learning network, judges text and the input of output
Whether the similarity between positive sample is greater than default similarity threshold, if the phase between the text and the positive sample of input of output
When being greater than default similarity threshold like degree, then terminate the training of the modal particle model based on deep learning network.
The similarity between the text of output and the positive sample of input, template can be calculated by the method for template matching
The method matched is the prior art, and in this not go into detail by the present invention.
Specifically, it is described by first matching result and default second text database match including：
1) word in first matching result is converted into the first phonetic；
2) judge in default second text database with the presence or absence of the second phonetic identical with first phonetic；
It 3), will when determining in default second text database in the presence of the second phonetic identical with first phonetic
The corresponding word of second phonetic extracts, as the corresponding word of the first phonetic.
For example, it is assumed that the first matching result is " this is an original megaseisms ", by the word in the first matching result
Being converted to the first phonetic is " zhe shi yige yuanshi juzhen "；Profession is stored in default second text database
Word " matrix " and corresponding second phonetic " juzheng ", then exist in determining default second text database with it is described
When identical second phonetic of the first phonetic, the corresponding word " matrix " of the second phonetic " juzheng " is extracted, as first
The corresponding word of phonetic " juzheng ", the second obtained matching result are " this is an original matrix ".
Specifically, it is described by second matching result and default third text database match including：Judge institute
It states in the second matching result with the presence or absence of the third word to match with the word in default third text database；When determining
When stating the third word for existing in the second matching result and matching with the word in default third text database, by described second
The third word to match in matching result is rejected.
S13, the speech recognition text rough draft according to the speech recognition text generation after the matching with editable state.
After being modified automatically to initial speech identification text according to matching result, the voice after being matched is known
Other text can first generate the speech recognition text rough draft with editable state.Refer to that user can with editable state
In the enterprising edlin operation of the speech recognition text rough draft of generation.The edit operation may include：Confirmation operation, modification behaviour
Make.
The confirmation operation refers to that user confirms that the speech recognition text rough draft is correct, that is, determines modified language
Sound identification text does not need to carry out any modification operation.The modification operation refers to that user confirms the speech recognition text rough draft
It is wrong, individual or a small amount of word needs are adjusted, i.e., modified speech recognition text also needs to carry out people again
Work manual modification.
S14, when detect have received edit operation on the speech recognition text rough draft when, according to the edit operation
Speech recognition text generation afterwards have can not editing mode speech recognition text, as final speech recognition text.
Edit operation whether is had received on the speech recognition text rough draft to be detected in the following manner：It detects
Whether touch operation is had received in the pre-set button surveyed on the speech recognition text rough draft, is connect in pre-set button when detecting
When having received touch operation, it is believed that detect and have received edit operation on the speech recognition text rough draft；When not detecting
When having received touch operation on to pre-set button, it is believed that detect and be not received by editor on the speech recognition text rough draft
Operation.
The pre-set button can be ACK button, can also be modification button.The button can be virtual icon, also
It can be physical button.The edit operation of corresponding ACK button is confirmation operation, and the edit operation of corresponding modification button is modification
Operation.
The speech recognition text generation according to after the edit operation have can not editing mode speech recognition text
Originally include：
When the edit operation received be confirmation operation when, directly generate with can not editing mode speech recognition text
This；
When the edit operation received is modification operation, receives the manual modification of user and saves the new content of modification,
When receiving confirmation operation again, generate have can not editing mode speech recognition text.
Preferably, after the edit operation for receiving user is modification operation, the method can also include：Correspondence is repaired
The original word and the modified neologism of user for changing place are associated storage；When subsequently through speech recognition technology, according to
Conference voice to be identified is converted text by modified neologism.
Pass through each modification of record user and by the original word at corresponding modification and the modified neologism of user into
Row associated storage is conducive to subsequent when carrying out speech recognition again, the neologism that can directly modify using user, to reduce knowledge
Other error rate improves the accuracy of speech recognition, the especially trouble of reduction user modification.
Preferably, in pre-set text database, the method can also include：It is corresponding more that each word is stored in advance
Kind form, the diversified forms may include, but be not limited to：Simplified and traditional body form plus space form and nearly word form etc..It is described by institute
It states initial speech and identifies that text match with pre-set text database further including：The environment according to locating for meeting, will be described first
Beginning speech recognition text diversified forms corresponding with each word of pre-set text database are matched, and obtain meeting meeting institute
Locate the speech recognition text of environment.
Environment locating for meeting may include：The participant of meeting, meeting hold ground.
By the corresponding diversified forms of setting word, and the environment according to locating for meeting is by initial speech identification text and in advance
It, can be sensitive to avoid some modal particles containing space, taboo if the word of every kind of form in text database is matched
Word can not identify.In addition it is possible to using in different occasions, for example, when the participant of meeting is Taiwanese, in Taiwan
In the case where etc. the complex form of Chinese characters accustomed to using, the initial speech that identifies identification text may be the complex form of Chinese characters and/or simplified Chinese character, thus
It is necessary to match the simplified and traditional body form of word, the speech recognition text for meeting the complex form of Chinese characters that Taiwanese gets used to is obtained
This.For another example, meeting is continent with holding, and according to the habit of mainlander, initial speech will be identified text and pre-set text number
After being matched according to the word of every kind of form in library, the speech recognition text for meeting the simplified Chinese character that mainlander gets used to is obtained
This.
It is of the present invention by meeting speech recognition be text method, pass through speech recognition technology for meeting to be identified
Voice is converted into text, identifies text as initial speech；By the initial speech identification text and pre-set text database into
Row matching, the speech recognition text after being matched；There is editable shape according to the speech recognition text generation after the matching
The speech recognition text rough draft of state；When detect have received edit operation on the speech recognition text rough draft when, according to institute
Speech recognition text generation after stating edit operation have can not editing mode speech recognition text, as final speech recognition
Text.After passing through ASR technology and being converted to Text Mode to the voice of meeting, by calling the content for having deposited dictionary in text
Appearance scans for, and the operation such as is replaced, deleted accordingly respectively.But it will lead to due to the minimum probability of operation such as replacing and deleting
It is the text of mistake by original correct text modification, so modified text can be in by system again with revisable mode
Now give user, while the place for marking out modified mistake confirms for user, user can place to system misoperation again
It modifies.I.e. by carrying out after tentatively identifying to voice to be identified, first time matching is carried out with pre-set text library, then pass through
It is artificial to carry out second of confirmation.Process can effectively ensure that the correctness of text output content twice, improve traditional voice
Turn the unreasonable place of literal expression in text, effectively reduces the proof-reading amount of conference content, improve efficiency.
The above is only a specific embodiment of the invention, but scope of protection of the present invention is not limited thereto, for
For those skilled in the art, without departing from the concept of the premise of the invention, improvement, but these can also be made
It all belongs to the scope of protection of the present invention.
Below with reference to the 2nd to 3 figure, respectively to the electronic equipment for realizing the above-mentioned method by meeting speech recognition for text
Functional module and hardware configuration are introduced.
Embodiment two
Fig. 2 is the present invention by the functional block diagram in device preferred embodiment that meeting speech recognition is text.
In some embodiments, described to run on the device 20 that meeting speech recognition is text in electronic equipment.It is described
It may include multiple functional modules as composed by program code segments by the device 20 that meeting speech recognition is text.It is described will
View speech recognition can store in memory for the program code of each program segment in the device 20 of text, and by least one
Performed by a processor, using execution (being detailed in Fig. 1 and its associated description) by meeting speech recognition as the function of text.
In the present embodiment, the electronic equipment by function of the device 20 according to performed by it that meeting speech recognition is text
Can, multiple functional modules can be divided into.The functional module may include：Identification module 201, generates matching module 202
Module 203, detecting module 204, relating module 205 and setup module 206.The so-called module of the present invention refers to that one kind can be by extremely
A few processor is performed and can complete the series of computation machine program segment of fixed function, and storage is in memory.
In some embodiments, it will be described in detail in subsequent embodiment about the function of each module.
Identification module 201, for converting text for conference voice to be identified by speech recognition technology, as initial
Speech recognition text.
In the present embodiment, the identification module 201 converts text for conference voice to be identified by speech recognition technology
This detailed process includes：
1) audio frequency characteristics for extracting conference voice to be identified, are converted to the acoustic feature vector of preset length；
2) described eigenvector is decoded by word order according to decoding algorithm；
3) the sub- word of corresponding word order is obtained by HMM phoneme model, the sub- word is initial consonant and simple or compound vowel of a Chinese syllable；
4) multiple sub- words are spliced by text according to preset Pronounceable dictionary；
5) it decodes to obtain optimal sequence using language model syntax rule, obtains text.
The syntax rule is viterbi algorithm.For example, the conference voice to be identified is " hello ", by spy
Sign is converted into the acoustic feature vectors of 39 dimensions after extracting, by multiple HMM phoneme models obtain corresponding multiple sub- word/n//i//
Multiple sub- words are spliced into word according to preset Pronounceable dictionary by h//ao/, as you, Buddhist nun；It is good, number.It is decoded by viterbi algorithm
Obtain optimal sequence " hello " and by text output.
Matching module 202 obtains for matching with pre-set text database initial speech identification text
Speech recognition text after matching.
In the present embodiment, at least three text databases can be preset, for example, the first text database, the second text
Database and third circumferential edge library.First text database can be exclusively used in storing multiple modal particles, such as " uh ", " ",
" right " etc., modal particle is unrelated with conference content, and easily influences voice again and be converted to the readability after text.Second text data
Library can be exclusively used in storing multiple professional words, such as " feature vector ", " eigenmatrix ", " tensor analysis ", and professional word is more complex,
It is thus easy batch during identifying voice and mistake occurs.It is quick that third text database can be exclusively used in storing multiple taboos
Feel word, such as political relevant name, porns, gambling and drugs, firearms and ammunition class, swears at people and satirizes class button invalid information class word at supertition heresy,
The appearance of taboo sensitive word be easy to cause adverse effect.The present invention can also preset the 4th text data according to the actual situation
Library etc. is exclusively used in the sentence of storage name or place name etc..Herein for pre-set circumferential edge library quantity and
Corresponding content is not specifically limited.
The matching module 202 carries out initial speech identification text to match specific packet with pre-set text database
It includes：
1) initial speech identification text is matched with default first text database, obtains the first matching knot
Fruit；
2) first matching result is matched with default second text database, obtains the second matching result；
3) second matching result is matched with default third text database, obtains third matching result.
Three independently operated threads can be preset：First thread, the second thread and third thread, first thread are held
It is about to the initial speech identification text to refer to the program that default first text database is matched to obtain the first matching result
It enables, first matching result is matched to obtain the second matching result by the execution of the second thread with default second text database
Program instruction, third thread, which executes, refers to second matching result and the matched program of default third text database progress
It enables.After first thread is finished, the second thread is executed at once, after the second thread is finished, executes third line at once
Journey, three independently operated threads of setting execute different program instructions respectively, are conducive to improve matched speed, save matching
Time.
In other embodiments, a thread can also be only set, and being sequentially carried out by one thread will be described initial
Speech recognition text is matched to obtain with default first text database the program instruction of the first matching result, by described first
Matching result is matched to obtain the program instruction of the second matching result, be matched described second with default second text database
As a result matched program instruction is carried out with default third text database.
The matching module 202 can also include：First matched sub-block 2020, the second matched sub-block 2022, third
Matched sub-block 2024.
Specifically, first matched sub-block 2020 by initial speech identification text and presets the first text data
Library carries out matching：Judge to whether there is in the initial speech identification text and the word in default first text database
The first word to match；Exist and the word in default first text database when determining in the initial speech identification text
When the first word to match, the initial speech is identified that the first word to match in text is handled.
Preferably, first matched sub-block 2020 is also used to：Deep learning net is based on according to training in advance
Whether the first word to match described in the modal particle model judgement of network is modal particle to be deleted；Match described in the determination
When first word is modal particle to be deleted, the initial speech is identified that the first word to match in text is rejected；
When the first word to match described in the determination is not modal particle to be deleted, the initial speech is identified and is matched in text
The first word retained.
For example, it is assumed that initial speech identifies that text is " this is pretty good ", deposits in default first text database
Modal particle " this " is contained, then phase has been determined after being matched initial speech identification text with default first text database
The word matched is " this ", then judges the phase according to the modal particle model based on deep learning network of training in advance
Whether matched first word " this " is modal particle to be deleted, and the modal particle model based on deep learning network determines
The first word " this " to match is not belonging to modal particle to be deleted in " this is pretty good ", then by the initial speech
The first word to match in identification text is retained, and the first obtained matching result is " this is pretty good ".
For another example, it is assumed that initial speech identifies that text is " this, we will have a meeting ", deposits in default first text database
Modal particle " this " is contained, then phase has been determined after being matched initial speech identification text with default first text database
The word matched is " this ", then judges the phase according to the modal particle model based on deep learning network of training in advance
Whether matched first word " this " is modal particle to be deleted, and the modal particle model based on deep learning network determines
The first word " this " to match belongs to modal particle to be deleted in " this, we will have a meeting ", then will be described initial
The first word to match in speech recognition text is rejected, and the first obtained matching result is " we will have a meeting ".
Preferably, modal particle model process of first matched sub-block 2020 training based on deep learning network can be with
Including：
1) text largely with the word in the first text database is obtained；
2) text is divided into positive sample text and negative sample text, the positive sample text is to need to retain the tone
The text of word, the negative sample text are the text for needing to delete modal particle；
For example, the word in the first text database is " this ", then it is available largely to contain " this " word
This paper of language, such as " this project is currently ongoing ", " whom this people is ", " this, uh, also in inquiry ", " this can
With such ", wherein " this " is the language for needing to retain in " this project is currently ongoing " and " whom this people is "
Gas word, " this " in " this, uh, also in inquiry ", " this can be such " is the modal particle for needing to delete.
3) the positive sample text is stamped into first identifier, the negative sample text is stamped into second identifier；
The word that the first identifier is used to identify in sample needs to retain, for example, it may be " 1 ".The second identifier
Word for identifying in sample needs to delete, for example, it may be " 0 ".
4) the positive sample text input is trained into deep learning network, judges text and the input of output
Whether the similarity between positive sample is greater than default similarity threshold, if the phase between the text and the positive sample of input of output
When being greater than default similarity threshold like degree, then terminate the training of the modal particle model based on deep learning network.
The similarity between the text of output and the positive sample of input, template can be calculated by the method for template matching
The method matched is the prior art, and in this not go into detail by the present invention.
Specifically, second matched sub-block 2022 is by described by first matching result and default second textual data
Carrying out matching according to library includes：Word in first matching result is converted into the first phonetic；Judge default second text
It whether there is the second phonetic identical with first phonetic in database；When determining in default second text database
When in the presence of the second phonetic identical with first phonetic, the corresponding word of the second phonetic is extracted, as the first phonetic
Corresponding word.
For example, it is assumed that the first matching result is " this is an original megaseisms ", by the word in the first matching result
Being converted to the first phonetic is " zhe shi yige yuanshi juzhen "；Profession is stored in default second text database
Word " matrix " and corresponding second phonetic " juzheng ", then exist in determining default second text database with it is described
When identical second phonetic of the first phonetic, the corresponding word " matrix " of the second phonetic " juzheng " is extracted, as first
The corresponding word of phonetic " juzheng ", the second obtained matching result are " this is an original matrix ".
Specifically, the third matched sub-block 2024 by second matching result and default third text database into
Row matches：Judge in second matching result with the presence or absence of matching with the word in default third text database
Third word；There is the third that matches with the word in default third text database in second matching result when determining
When word, the third word to match in second matching result is rejected.
Generation module 203, for having the voice of editable state according to the speech recognition text generation after the matching
Identify text rough draft.
After being modified automatically to initial speech identification text according to matching result, the voice after being matched is known
Other text can first generate the speech recognition text rough draft with editable state.Refer to that user can with editable state
In the enterprising edlin operation of the speech recognition text rough draft of generation.The edit operation may include：Confirmation operation, modification behaviour
Make.
The confirmation operation refers to that user confirms that the speech recognition text rough draft is correct, that is, determines modified language
Sound identification text does not need to carry out any modification operation.The modification operation refers to that user confirms the speech recognition text rough draft
It is wrong, individual or a small amount of word needs are adjusted, i.e., modified speech recognition text also needs to carry out people again
Work manual modification.
Whether detecting module 204 has received edit operation on the speech recognition text rough draft for detecting.
Whether having received edit operation on 204 detecting voice of the detecting module identification text rough draft includes：Detecting institute
It states and whether has received touch operation in the pre-set button on speech recognition text rough draft, received in pre-set button when detecting
When touch operation, it is believed that detect and have received edit operation on the speech recognition text rough draft；It is pre- when not detecting
If have received touch operation on button, it is believed that detect and be not received by editor behaviour on the speech recognition text rough draft
Make.
The pre-set button can be ACK button, can also be modification button.The button can be virtual icon, also
It can be physical button.The edit operation of corresponding ACK button is confirmation operation, and the edit operation of corresponding modification button is modification
Operation.
The generation module 203 is also used to receive on the speech recognition text rough draft when the detecting module detects
When having arrived edit operation, according to the speech recognition text generation after the edit operation have can not editing mode speech recognition
Text, as final speech recognition text.
The generation module 203 has according to speech recognition text generation after the edit operation can not editing mode
Speech recognition text includes：
When the edit operation received be confirmation operation when, directly generate with can not editing mode speech recognition text
This；
When the edit operation received is modification operation, receives the manual modification of user and saves the new content of modification,
When receiving confirmation operation again, generate have can not editing mode speech recognition text.
Preferably, after the edit operation for receiving user is modification operation, described by meeting speech recognition is text
Device 20 can also include：Relating module 205, for will correspond to original word at modification and the modified neologism of user into
Row associated storage；The identification module 201 is also used to when subsequently through speech recognition technology, will according to modified neologism
Conference voice to be identified is converted into text.
Pass through each modification of record user and by the original word at corresponding modification and the modified neologism of user into
Row associated storage is conducive to subsequent when carrying out speech recognition again, the neologism that can directly modify using user, to reduce knowledge
Other error rate improves the accuracy of speech recognition, the especially trouble of reduction user modification.
Preferably, described to include by the device 20 that meeting speech recognition is text in pre-set text database：
Setup module 206, for the corresponding diversified forms of each word to be stored in advance, the diversified forms may include, but be not limited to：
Simplified and traditional body form plus space form and nearly word form etc..
It is described that the initial speech is identified that text match with pre-set text database further including：According to locating for meeting
Environment, by corresponding with each word of the pre-set text database diversified forms progress of initial speech identification text
Match, obtains the speech recognition text for meeting meeting local environment.
Environment locating for meeting may include：The participant of meeting, meeting hold ground.
By the corresponding diversified forms of setting word, and the environment according to locating for meeting is by initial speech identification text and in advance
It, can be sensitive to avoid some modal particles containing space, taboo if the word of every kind of form in text database is matched
Word can not identify.In addition it is possible to using in different occasions, for example, when the participant of meeting is Taiwanese, in Taiwan
In the case where etc. the complex form of Chinese characters accustomed to using, the initial speech that identifies identification text may be the complex form of Chinese characters and/or simplified Chinese character, thus
It is necessary to match the simplified and traditional body form of word, the speech recognition text for meeting the complex form of Chinese characters that Taiwanese gets used to is obtained
This.For another example, meeting is continent with holding, and according to the habit of mainlander, initial speech will be identified text and pre-set text number
After being matched according to the word of every kind of form in library, the speech recognition text for meeting the simplified Chinese character that mainlander gets used to is obtained
This.
It is of the present invention by meeting speech recognition be text device, pass through speech recognition technology for meeting to be identified
Voice is converted into text, identifies text as initial speech；By the initial speech identification text and pre-set text database into
Row matching, the speech recognition text after being matched；There is editable shape according to the speech recognition text generation after the matching
The speech recognition text rough draft of state；When detect have received edit operation on the speech recognition text rough draft when, according to institute
Speech recognition text generation after stating edit operation have can not editing mode speech recognition text, as final speech recognition
Text.After passing through ASR technology and being converted to Text Mode to the voice of meeting, by calling the content for having deposited dictionary in text
Appearance scans for, and the operation such as is replaced, deleted accordingly respectively.But it will lead to due to the minimum probability of operation such as replacing and deleting
It is the text of mistake by original correct text modification, so modified text can be in by system again with revisable mode
Now give user, while the place for marking out modified mistake confirms for user, user can place to system misoperation again
It modifies.I.e. by carrying out after tentatively identifying to voice to be identified, first time matching is carried out with pre-set text library, then pass through
It is artificial to carry out second of confirmation.Process can effectively ensure that the correctness of text output content twice, improve traditional voice
Turn the unreasonable place of literal expression in text, effectively reduces the proof-reading amount of conference content, improve efficiency.
The above-mentioned integrated unit realized in the form of software function module, can store and computer-readable deposit at one
In storage media.Above-mentioned software function module is stored in a storage medium, including some instructions are used so that a computer
It is each that equipment (can be personal computer, double screen equipment or the network equipment etc.) or processor (processor) execute the present invention
The part of a embodiment the method.
Embodiment three
Fig. 3 is the schematic diagram for the electronic equipment that the embodiment of the present invention five provides.
The electronic equipment 3 includes：Memory 31, at least one processor 32 are stored in the memory 31 and can
The computer program 33 and at least one communication bus 34 run at least one described processor 32.
At least one described processor 32 realized when executing the computer program 33 it is above-mentioned by meeting speech recognition for text
Step in this embodiment of the method.
Illustratively, the computer program 33 can be divided into one or more module/units, it is one or
Multiple module/units are stored in the memory 31, and are executed by least one described processor 32, to complete this hair
It is bright.One or more of module/units can be the series of computation machine program instruction section that can complete specific function, this refers to
Enable section for describing implementation procedure of the computer program 33 in the electronic equipment 3.
The electronic equipment 3 can be the calculating such as desktop PC, notebook, palm PC and cloud server and set
It is standby.It will be understood by those skilled in the art that the schematic diagram 3 is only the example of electronic equipment 3, do not constitute to electronic equipment
3 restriction may include perhaps combining certain components or different components, such as institute than illustrating more or fewer components
Stating electronic equipment 3 can also include input-output equipment, network access equipment, bus etc..
At least one described processor 32 can be central processing unit (Central Processing Unit, CPU),
It can also be other general processors, digital signal processor (Digital Signal Processor, DSP), dedicated integrated
Circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..The processor 32 can be microprocessor or the processor 32 is also possible to any conventional processor
Deng the processor 32 is the control centre of the electronic equipment 3, utilizes various interfaces and the entire electronic equipment 3 of connection
Various pieces.
The memory 31 can be used for storing the computer program 33 and/or module/unit, and the processor 32 passes through
Operation executes the computer program and/or module/unit being stored in the memory 31, and calls and be stored in memory
Data in 31 realize the various functions of the electronic equipment 3.The memory 31 can mainly include storing program area and storage
Data field, wherein storing program area can application program needed for storage program area, at least one function (for example sound plays
Function, image player function etc.) etc.；Storage data area, which can be stored, uses created data (such as sound according to electronic equipment 3
Frequency evidence, phone directory etc.) etc..In addition, memory 31 may include high-speed random access memory, it can also include non-volatile
Memory, such as hard disk, memory, plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital
(Secure Digital, SD) card, flash card (Flash Card), at least one disk memory, flush memory device or other
Volatile solid-state part.
If the integrated module/unit of the electronic equipment 3 is realized in the form of SFU software functional unit and as independent
Product when selling or using, can store in a computer readable storage medium.Based on this understanding, the present invention is real
All or part of the process in existing above-described embodiment method, can also instruct relevant hardware come complete by computer program
At the computer program can be stored in a computer readable storage medium, which is being executed by processor
When, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program code, described
Computer program code can be source code form, object identification code form, executable file or certain intermediate forms etc..The meter
Calculation machine readable medium may include：Any entity or recording medium, USB flash disk, the movement of the computer program code can be carried
Hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), random access memory
(RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..It needs to illustrate
It is that the content that the computer-readable medium includes can be fitted according to the requirement made laws in jurisdiction with patent practice
When increase and decrease, such as in certain jurisdictions, according to legislation and patent practice, computer-readable medium does not include electric carrier wave letter
Number and telecommunication signal.
In several embodiments provided by the present invention, it should be understood that disclosed electronic equipment and method, Ke Yitong
Other modes are crossed to realize.For example, electronic equipment embodiment described above is only schematical, for example, the unit
Division, only a kind of logical function partition, there may be another division manner in actual implementation.
It, can also be in addition, each functional unit in each embodiment of the present invention can integrate in same treatment unit
It is that each unit physically exists alone, can also be integrated in same unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of hardware adds software function module.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included in the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.This
Outside, it is clear that one word of " comprising " is not excluded for other units or, odd number is not excluded for plural number.The multiple units stated in system claims
Or device can also be implemented through software or hardware by a unit or device.The first, the second equal words are used to indicate name
Claim, and does not indicate any particular order.
Finally it should be noted that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although reference
Preferred embodiment describes the invention in detail, those skilled in the art should understand that, it can be to of the invention
Technical solution is modified or equivalent replacement, without departing from the spirit of the technical scheme of the invention range.
Claims (10)
1. it is a kind of by meeting speech recognition be text method, which is characterized in that the method includes：
Text is converted by conference voice to be identified by speech recognition technology, identifies text as initial speech；
Initial speech identification text is matched with pre-set text database, the speech recognition text after being matched；
There is the speech recognition text rough draft of editable state according to the speech recognition text generation after the matching；
When detect have received edit operation on the speech recognition text rough draft when, according to the voice after the edit operation
Identify text generation have can not editing mode speech recognition text, as final speech recognition text.
2. the method as described in claim 1, which is characterized in that described by initial speech identification text and pre-set text number
Carrying out matching according to library includes：
Initial speech identification text is matched with default first text database, obtains the first matching result；
First matching result is matched with default second text database, obtains the second matching result；
Second matching result is matched with default third text database；
Wherein, it is stored with multiple modal particles in default first text database, is deposited in default second text database
Multiple professional words and corresponding phonetic are contained, is stored with multiple taboo sensitive words in the default third text database.
3. method according to claim 2, which is characterized in that described by initial speech identification text and default first text
Database carries out matching：
Judge in initial speech identification text with the presence or absence of the to match with the word in default first text database
One word；
When determining in initial speech identification text there is first to match with the word in default first text database
When word, whether the first word to match according to the modal particle model judgement based on deep learning network of training in advance
For modal particle to be deleted；
When the first word to match described in the determination is modal particle to be deleted, the initial speech is identified into phase in text
The first word matched is rejected；
When the first word to match described in the determination is not modal particle to be deleted, the initial speech is identified into phase in text
Matched first word is retained.
4. method according to claim 2, which is characterized in that described by first matching result and default second textual data
Carrying out matching according to library includes：
Word in first matching result is converted into the first phonetic；
Judge in default second text database with the presence or absence of the second phonetic identical with first phonetic；
When determining in default second text database in the presence of the second phonetic identical with first phonetic, second is spelled
The corresponding word of sound extracts, as the corresponding word of the first phonetic.
5. method according to claim 2, which is characterized in that described by second matching result and default third textual data
Carrying out matching according to library includes：
Judge in second matching result with the presence or absence of the third word to match with the word in default third text database
Language；
There is the third word that matches with the word in default third text database in second matching result when determining
When, the third word to match in second matching result is rejected.
6. the method as described in claim 1, which is characterized in that described raw according to the speech recognition text after the edit operation
At have can not the speech recognition text of editing mode include：
When the edit operation received be confirmation operation when, directly generate with can not editing mode speech recognition text；
When the edit operation received is modification operation, receives the manual modification of user and save the new content of modification, when again
It is secondary when receiving confirmation operation, generate have can not editing mode speech recognition text.
7. the method as described in claim 1 to 6 any one, which is characterized in that the method also includes：
By at corresponding modification original word and the modified neologism of user be associated storage；
When subsequently through speech recognition technology, text is converted for conference voice to be identified according to modified neologism.
8. the method as described in claim 1 to 6 any one, which is characterized in that the method also includes：
The corresponding diversified forms of each word are stored in advance, the diversified forms include：Simplified and traditional body form plus space form and shape
Nearly word；
Initial speech identification text diversified forms corresponding with each word of pre-set text database are matched.
9. a kind of electronic equipment, which is characterized in that the electronic equipment includes processor and memory, and the processor is for holding
It realizes when the computer program stored in the row memory as claimed in any of claims 1 to 8 in one of claims by conference voice
The method for being identified as text.
10. a kind of computer readable storage medium, computer program, feature are stored on the computer readable storage medium
It is, realizes when the computer program is executed by processor as claimed in any of claims 1 to 8 in one of claims by conference voice
The method for being identified as text.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810581922.4A CN108847241B (en) | 2018-06-07 | 2018-06-07 | Method for recognizing conference voice as text, electronic device and storage medium |
| PCT/CN2018/108113 WO2019232991A1 (en) | 2018-06-07 | 2018-09-27 | Method for recognizing conference voice as text, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810581922.4A CN108847241B (en) | 2018-06-07 | 2018-06-07 | Method for recognizing conference voice as text, electronic device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108847241A true CN108847241A (en) | 2018-11-20 |
| CN108847241B CN108847241B (en) | 2022-09-13 |
Family
ID=64211364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810581922.4A Active CN108847241B (en) | 2018-06-07 | 2018-06-07 | Method for recognizing conference voice as text, electronic device and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN108847241B (en) |
| WO (1) | WO2019232991A1 (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109708256A (en) * | 2018-12-06 | 2019-05-03 | 珠海格力电器股份有限公司 | Voice determination method and device, storage medium and air conditioner |
| CN110335509A (en) * | 2019-07-09 | 2019-10-15 | 南阳理工学院 | A kind of primary school teaching apparatus for demonstrating |
| CN110459224A (en) * | 2019-07-31 | 2019-11-15 | 北京百度网讯科技有限公司 | Speech recognition result processing method, device, computer equipment and storage medium |
| CN110619879A (en) * | 2019-08-29 | 2019-12-27 | 深圳市梦网科技发展有限公司 | Voice recognition method and device |
| CN110969026A (en) * | 2019-11-27 | 2020-04-07 | 北京欧珀通信有限公司 | Translation output method and device, electronic equipment and storage medium |
| CN111177353A (en) * | 2019-12-27 | 2020-05-19 | 拉克诺德(深圳)科技有限公司 | Text record generation method and device, computer equipment and storage medium |
| CN111651960A (en) * | 2020-06-01 | 2020-09-11 | 杭州尚尚签网络科技有限公司 | Optical character joint training and recognition method for moving from contract simplified form to traditional form |
| CN111710328A (en) * | 2020-06-16 | 2020-09-25 | 北京爱医声科技有限公司 | Method, device and medium for selecting training samples of voice recognition model |
| CN112037792A (en) * | 2020-08-20 | 2020-12-04 | 北京字节跳动网络技术有限公司 | Voice recognition method and device, electronic equipment and storage medium |
| CN112468665A (en) * | 2020-11-05 | 2021-03-09 | 中国建设银行股份有限公司 | Method, device, equipment and storage medium for generating conference summary |
| CN112951210A (en) * | 2021-02-02 | 2021-06-11 | 虫洞创新平台(深圳)有限公司 | Speech recognition method and device, equipment and computer readable storage medium |
| CN113360623A (en) * | 2021-06-25 | 2021-09-07 | 达闼机器人有限公司 | Text matching method, electronic device and readable storage medium |
| CN113470644A (en) * | 2021-06-29 | 2021-10-01 | 读书郎教育科技有限公司 | Intelligent voice learning method and device based on voice recognition |
| WO2022037600A1 (en) * | 2020-08-18 | 2022-02-24 | 深圳市万普拉斯科技有限公司 | Abstract recording method and apparatus, and computer device and storage medium |
| CN114120977A (en) * | 2021-11-23 | 2022-03-01 | 四川虹美智能科技有限公司 | Speech recognition new word self-learning method and device |
| US11303464B2 (en) * | 2019-12-05 | 2022-04-12 | Microsoft Technology Licensing, Llc | Associating content items with images captured of meeting content |
| CN114822527A (en) * | 2021-10-11 | 2022-07-29 | 北京中电慧声科技有限公司 | Error correction method and device for converting voice into text, electronic equipment and storage medium |
| CN115623134A (en) * | 2022-10-08 | 2023-01-17 | 中国电信股份有限公司 | Conference audio processing method, device, equipment and storage medium |
| CN111368506B (en) * | 2018-12-24 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Text processing method and device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114666618B (en) * | 2022-03-15 | 2023-10-13 | 广州欢城文化传媒有限公司 | Audio auditing method, device, equipment and readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120143605A1 (en) * | 2010-12-01 | 2012-06-07 | Cisco Technology, Inc. | Conference transcription based on conference data |
| CN102956231A (en) * | 2011-08-23 | 2013-03-06 | 上海交通大学 | Voice key information recording device and method based on semi-automatic correction |
| CN104679729A (en) * | 2015-02-13 | 2015-06-03 | 广州市讯飞樽鸿信息技术有限公司 | Recorded message effective processing method and system |
| CN105206274A (en) * | 2015-10-30 | 2015-12-30 | 北京奇艺世纪科技有限公司 | Voice recognition post-processing method and device as well as voice recognition system |
| CN105206272A (en) * | 2015-09-06 | 2015-12-30 | 上海智臻智能网络科技股份有限公司 | Voice transmission control method and system |
| CN107590121A (en) * | 2016-07-08 | 2018-01-16 | 科大讯飞股份有限公司 | Text-normalization method and system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7668718B2 (en) * | 2001-07-17 | 2010-02-23 | Custom Speech Usa, Inc. | Synchronized pattern recognition source data processed by manual or automatic means for creation of shared speaker-dependent speech user profile |
| CN103902629B (en) * | 2012-12-28 | 2017-09-29 | 联想(北京)有限公司 | The electronic equipment and method for operating help are provided using voice |
| CN105976818B (en) * | 2016-04-26 | 2020-12-25 | Tcl科技集团股份有限公司 | Instruction recognition processing method and device |
-
2018
- 2018-06-07 CN CN201810581922.4A patent/CN108847241B/en active Active
- 2018-09-27 WO PCT/CN2018/108113 patent/WO2019232991A1/en active Application Filing
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120143605A1 (en) * | 2010-12-01 | 2012-06-07 | Cisco Technology, Inc. | Conference transcription based on conference data |
| CN102956231A (en) * | 2011-08-23 | 2013-03-06 | 上海交通大学 | Voice key information recording device and method based on semi-automatic correction |
| CN104679729A (en) * | 2015-02-13 | 2015-06-03 | 广州市讯飞樽鸿信息技术有限公司 | Recorded message effective processing method and system |
| CN105206272A (en) * | 2015-09-06 | 2015-12-30 | 上海智臻智能网络科技股份有限公司 | Voice transmission control method and system |
| CN105206274A (en) * | 2015-10-30 | 2015-12-30 | 北京奇艺世纪科技有限公司 | Voice recognition post-processing method and device as well as voice recognition system |
| CN107590121A (en) * | 2016-07-08 | 2018-01-16 | 科大讯飞股份有限公司 | Text-normalization method and system |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109708256A (en) * | 2018-12-06 | 2019-05-03 | 珠海格力电器股份有限公司 | Voice determination method and device, storage medium and air conditioner |
| CN109708256B (en) * | 2018-12-06 | 2020-07-03 | 珠海格力电器股份有限公司 | Voice determination method and device, storage medium and air conditioner |
| CN111368506B (en) * | 2018-12-24 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Text processing method and device |
| CN110335509A (en) * | 2019-07-09 | 2019-10-15 | 南阳理工学院 | A kind of primary school teaching apparatus for demonstrating |
| CN110459224A (en) * | 2019-07-31 | 2019-11-15 | 北京百度网讯科技有限公司 | Speech recognition result processing method, device, computer equipment and storage medium |
| CN110459224B (en) * | 2019-07-31 | 2022-02-25 | 北京百度网讯科技有限公司 | Speech recognition result processing method and device, computer equipment and storage medium |
| CN110619879A (en) * | 2019-08-29 | 2019-12-27 | 深圳市梦网科技发展有限公司 | Voice recognition method and device |
| CN110969026A (en) * | 2019-11-27 | 2020-04-07 | 北京欧珀通信有限公司 | Translation output method and device, electronic equipment and storage medium |
| US11303464B2 (en) * | 2019-12-05 | 2022-04-12 | Microsoft Technology Licensing, Llc | Associating content items with images captured of meeting content |
| CN111177353A (en) * | 2019-12-27 | 2020-05-19 | 拉克诺德(深圳)科技有限公司 | Text record generation method and device, computer equipment and storage medium |
| CN111651960A (en) * | 2020-06-01 | 2020-09-11 | 杭州尚尚签网络科技有限公司 | Optical character joint training and recognition method for moving from contract simplified form to traditional form |
| CN111651960B (en) * | 2020-06-01 | 2023-05-30 | 杭州尚尚签网络科技有限公司 | Optical character joint training and recognition method for transferring contract simplified body to complex body |
| CN111710328A (en) * | 2020-06-16 | 2020-09-25 | 北京爱医声科技有限公司 | Method, device and medium for selecting training samples of voice recognition model |
| CN111710328B (en) * | 2020-06-16 | 2024-01-12 | 北京爱医声科技有限公司 | Training sample selection method, device and medium for speech recognition model |
| WO2022037600A1 (en) * | 2020-08-18 | 2022-02-24 | 深圳市万普拉斯科技有限公司 | Abstract recording method and apparatus, and computer device and storage medium |
| CN112037792A (en) * | 2020-08-20 | 2020-12-04 | 北京字节跳动网络技术有限公司 | Voice recognition method and device, electronic equipment and storage medium |
| CN112468665A (en) * | 2020-11-05 | 2021-03-09 | 中国建设银行股份有限公司 | Method, device, equipment and storage medium for generating conference summary |
| CN112951210A (en) * | 2021-02-02 | 2021-06-11 | 虫洞创新平台(深圳)有限公司 | Speech recognition method and device, equipment and computer readable storage medium |
| CN113360623A (en) * | 2021-06-25 | 2021-09-07 | 达闼机器人有限公司 | Text matching method, electronic device and readable storage medium |
| CN113470644B (en) * | 2021-06-29 | 2023-09-26 | 读书郎教育科技有限公司 | Intelligent voice learning method and device based on voice recognition |
| CN113470644A (en) * | 2021-06-29 | 2021-10-01 | 读书郎教育科技有限公司 | Intelligent voice learning method and device based on voice recognition |
| CN114822527A (en) * | 2021-10-11 | 2022-07-29 | 北京中电慧声科技有限公司 | Error correction method and device for converting voice into text, electronic equipment and storage medium |
| CN114120977A (en) * | 2021-11-23 | 2022-03-01 | 四川虹美智能科技有限公司 | Speech recognition new word self-learning method and device |
| CN115623134A (en) * | 2022-10-08 | 2023-01-17 | 中国电信股份有限公司 | Conference audio processing method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2019232991A1 (en) | 2019-12-12 |
| CN108847241B (en) | 2022-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108847241A (en) | It is method, electronic equipment and the storage medium of text by meeting speech recognition | |
| CN110491382B (en) | Speech recognition method and device based on artificial intelligence and speech interaction equipment | |
| CN109918680B (en) | Entity identification method and device and computer equipment | |
| CN109887497B (en) | Modeling method, device and equipment for speech recognition | |
| CN110110041A (en) | Wrong word correcting method, device, computer installation and storage medium | |
| CN110473566A (en) | Audio separation method, device, electronic equipment and computer readable storage medium | |
| CN108549637A (en) | Method for recognizing semantics, device based on phonetic and interactive system | |
| CN113836277A (en) | Machine learning system for digital assistant | |
| EP3405912A1 (en) | Analyzing textual data | |
| CN109523989A (en) | Phoneme synthesizing method, speech synthetic device, storage medium and electronic equipment | |
| CN107220235A (en) | Speech recognition error correction method, device and storage medium based on artificial intelligence | |
| CN114580382A (en) | Text error correction method and device | |
| CN108682420A (en) | A kind of voice and video telephone accent recognition method and terminal device | |
| CN112992125B (en) | Voice recognition method and device, electronic equipment and readable storage medium | |
| CN109448704A (en) | Construction method, device, server and the storage medium of tone decoding figure | |
| CN111145720A (en) | Method, system, device and storage medium for converting text into voice | |
| CN110473527B (en) | Method and system for voice recognition | |
| WO2002029783A1 (en) | Method and system for using rule-based knowledge to build a class-based domain specific statistical language model | |
| CN112463942A (en) | Text processing method and device, electronic equipment and computer readable storage medium | |
| CN114625855A (en) | Method, apparatus, device and medium for generating dialogue information | |
| CN114999463B (en) | Voice recognition method, device, equipment and medium | |
| CN110852075B (en) | Voice transcription method and device capable of automatically adding punctuation marks and readable storage medium | |
| CN114783424A (en) | Text corpus screening method, device, equipment and storage medium | |
| CN110020429A (en) | Method for recognizing semantics and equipment | |
| Bangalore et al. | Balancing data-driven and rule-based approaches in the context of a multimodal conversational system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |