CN108932945A - A kind of processing method and processing device of phonetic order - Google Patents
A kind of processing method and processing device of phonetic order Download PDFInfo
- Publication number
- CN108932945A CN108932945A CN201810233853.8A CN201810233853A CN108932945A CN 108932945 A CN108932945 A CN 108932945A CN 201810233853 A CN201810233853 A CN 201810233853A CN 108932945 A CN108932945 A CN 108932945A
- Authority
- CN
- China
- Prior art keywords
- phonetic order
- parsing
- error sample
- resource
- intended
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 29
- 238000012545 processing Methods 0.000 title claims abstract description 17
- 238000000034 method Methods 0.000 claims abstract description 50
- 238000001514 detection method Methods 0.000 claims description 37
- 238000003066 decision tree Methods 0.000 claims description 36
- 238000003860 storage Methods 0.000 claims description 10
- 238000004321 preservation Methods 0.000 claims description 4
- 238000007689 inspection Methods 0.000 claims 1
- 238000006243 chemical reaction Methods 0.000 abstract description 6
- 230000003993 interaction Effects 0.000 abstract description 6
- 238000004422 calculation algorithm Methods 0.000 description 16
- 230000002354 daily effect Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 12
- 235000013550 pizza Nutrition 0.000 description 11
- 235000011475 lollipops Nutrition 0.000 description 10
- 238000007637 random forest analysis Methods 0.000 description 10
- 230000002452 interceptive effect Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 230000009471 action Effects 0.000 description 7
- 230000010056 antibody-dependent cellular cytotoxicity Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000012216 screening Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000005194 fractionation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000005303 weighing 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
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L2015/223—Execution procedure of a spoken command
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
Abstract
This application discloses a kind of processing method and processing device of phonetic order, the method includes：Receive the phonetic order for coming that self terminal includes user's original intent；Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；The text information is parsed, determines that parsing corresponding to the text information is intended to；It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent to the terminal；It determines that the original intent is unmet, the phonetic order, text information and parsing intention is labeled as error sample and saved to error sample library.The method decides whether to obtain error sample according to the reaction pattern of human-computer interaction, carries out the operation of error label manually without user, increase a possibility that getting error sample.
Description
Technical field
This application involves interactive voice technical field of intelligent equipment, a kind of processing method more particularly to phonetic order and
Device.
Background technique
With the development of artificial intelligence technology, occur smart machine various in style in the market, common are intelligence and set
Standby includes smart phone, intelligent sound box, smart television, intelligent robot etc..In order to promote the usage experience of user, many intelligence
Equipment all provides the function of voice input and voice output.The language that the voice interactive system of these smart machines is inputted according to user
Sound instructs the intention for determining user, to provide various services for user.
In common voice interactive system, it is generally divided into three parts and is handled come the instruction inputted to user.First
The phonetic order that user inputs is converted into text by speech recognition system ASR (Automatic speech recognition)
Word；Then the intention as representated by semantic resolution system NLP (Natural language processing) parsing text；Most
Realizing and being intended to complete for task is executed by requesting various resources afterwards.
Wherein, speech recognition system and natural language processing system require a large amount of labeled data to be trained.?
After error sample detection is online, it is also necessary to constantly manually be marked to user's input, to improve error sample detection model
Accuracy.It is the autonomous marking error sample of man-machine interaction mode for passing through active by user mostly in the prior art.Due to
There is no mode standard, corresponding same original intents might have various, multifarious for the phonetic order of user's input
Phonetic order collects a large amount of labeled data, especially those labeled data being erroneously identified, and can usually examine to error sample
The performance for surveying model brings biggish raising.But it in the prior art, needs user to be switched in other interactive systems actively to carry out
The mark of error sample, because of the cumbersome movement for causing most of user to abandon completing data mark, actually very
Difficulty gets the data that user actively submits, it has to expend a large amount of manpower and material resources and collect error sample deposit error sample library
In, this causes the training error sample detection model in existing voice interactive system to cannot achieve quickly mentioning for user experience
It rises.
Summary of the invention
In order to solve the problems in the existing technology, the embodiment of the present application provides a kind of processing side of phonetic order
Method, device, smart machine and computer readable storage medium, to solve to obtain error label sample automatically from man-machine interactive operation
This problem of, to quickly improve the user experience of voice interactive system.
On the one hand the embodiment of the present application provides a kind of processing method of phonetic order, the method includes：
Receive the phonetic order for coming that self terminal includes user's original intent；
Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；
The text information is parsed, determines that parsing corresponding to the text information is intended to；
It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent to described
Terminal；
It determines that the original intent is unmet, is intended to the phonetic order, text information and parsing to be labeled as mistake
Sample preservation is missed to error sample library.
Optionally, the determination original intent is unmet includes：
The parsing for repeating to receive same user in preset period of time is intended to identical phonetic order；Or
It receives and is intended to the phonetic order, text information and parsing from the terminal to be labeled as the letter of error sample
Breath.
Optionally, determine whether to repeat to receive in preset period of time the parsing of same user in a manner of decision tree
It is intended to identical phonetic order.
Optionally, the method also includes：
Save the record of resource retrieval corresponding with the parsing intention；
If determining that the parsing is intended to unmet reason according to preset rules based on the retrieval record is money
Caused by the retrieval of source, then the error sample is rejected from the error sample library.
Optionally, the method also includes：
Save the record of resource retrieval corresponding with the parsing intention；
The matching degree that the resource retrieval and the parsing are intended to is calculated, and the value for the matching degree being calculated is stored in
In retrieval record；
The value of the matching degree is less than preset threshold, then rejects the error sample from the error sample library.
On the other hand the embodiment of the present application also provides a kind of processing method of phonetic order, the method includes：
The phonetic order is simultaneously sent to server by the phonetic order of original intent of the acquisition comprising user；
Resource needed for executing the phonetic order is obtained from server end；
It determines that the original intent is unmet, send to the server by the phonetic order and is based on institute's predicate
The parsing that sound instruction carries out the text information of speech recognition acquisition and carries out parsing acquisition to the text information is intended to mark
For the information of error sample.
Optionally, after obtaining resource required for executing the phonetic order, the method also includes：
The information of the corresponding execution movement of the phonetic order is provided based on the resource got；
The determination original intent is unmet to include：
Capture the instruction for abandoning executing the corresponding execution movement of the phonetic order.
Optionally, after obtaining resource required for executing the phonetic order, the method also includes：
Execute that the phonetic order is corresponding to execute movement based on the resource that gets；
The determination original intent is unmet to include：
Capture the instruction that the corresponding execution movement of the phonetic order is terminated in scheduled time threshold value.
On the other hand the embodiment of the present application also provides a kind of processing unit of phonetic order, described device includes：Receive mould
Block, speech recognition module, parsing module, resource retrieval module, the first error sample detection module and error sample library；Wherein,
The receiving module is configured as receiving the phonetic order that self terminal includes user's original intent；The speech recognition module quilt
It is configured to carry out speech recognition to the phonetic order, generates the text information of the phonetic order；The parsing module is matched
It is set to and the text information is parsed, determine that parsing corresponding to the text information is intended to；The resource retrieval module
It is configured as being intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent to described
Terminal；The first error sample detection module is configured to determine that the original intent is unmet, and the voice is referred to
It enables, text information and parsing intention are labeled as error sample and save to error sample library；The error sample library is configured as depositing
Store up the error sample.
Optionally, the first error sample detection module determines that the original intent is unmet and includes：
The parsing for repeating to receive same user in preset period of time is intended to identical phonetic order；Or
It receives and is intended to the phonetic order, text information and parsing from the terminal to be labeled as the letter of error sample
Breath.
Optionally, the first error sample detection module is configured as determining whether in a manner of decision tree when default
Between repeat to receive the parsing of same user in the period and be intended to identical phonetic order.
Optionally, the first error sample detection module is additionally configured to：
Save the record of resource retrieval corresponding with the parsing intention；
If determining that the parsing is intended to unmet reason according to preset rules based on the retrieval record is money
Caused by the retrieval of source, then the error sample is rejected from the error sample library.
Optionally, the first error sample detection module is additionally configured to：
Save the record of resource retrieval corresponding with the parsing intention；
The matching degree that the resource retrieval and the parsing are intended to is calculated, and the value for the matching degree being calculated is stored in
In retrieval record；
The value of the matching degree is less than preset threshold, then rejects the error sample from the error sample library.
Optionally, described device includes：Acquisition module, execution module and the second error sample detection module；The acquisition
Module is configured as the phonetic order of original intent of the acquisition comprising user and the phonetic order is sent to server；It is described
Execution module is configured as resource needed for obtaining the execution phonetic order from server end；The second error sample detection
Module is configured to determine that the original intent is unmet, sends to the server by the phonetic order and is based on institute
The parsing that phonetic order carries out the text information of speech recognition acquisition and carries out parsing acquisition to the text information is stated to be intended to
It is labeled as the information of error sample.
Optionally, the execution module is additionally configured to after obtaining resource required for executing the phonetic order,
The information of the corresponding execution movement of the phonetic order is provided based on the resource got；
The determination original intent is unmet to include：
Capture the instruction for abandoning executing the corresponding execution movement of the phonetic order.
Optionally, the execution module is additionally configured to after obtaining resource required for executing the phonetic order, is based on
The resource execution phonetic order got is corresponding to execute movement；
The determination original intent is unmet to include：
Capture the instruction that the corresponding execution movement of the phonetic order is terminated in scheduled time threshold value.
On the other hand the embodiment of the present application also provides a kind of smart machine, including memory, processor and be stored in storage
On device and the computer instruction that can run on a processor, which is characterized in that the processor is realized such as when executing described instruction
The processing method of the upper phonetic order.
On the other hand embodiments herein also provides a kind of computer readable storage medium, be stored thereon with computer and refer to
It enables, which is characterized in that the processing method of phonetic order as described above is realized in the instruction when being executed by processor.
The processing method and processing device of phonetic order provided by the present application can obtain automatically those from man-machine interactive operation
The data being erroneously identified label it as error sample and save into error sample library, and it is wrong that this not only greatly reduces mark
Accidentally human cost consumed by sample, and the optimization efficiency of error sample detection model is significantly improved, and then be effectively improved
The user experience of voice interactive system.
Detailed description of the invention
Fig. 1 is the flow diagram of the processing method of the phonetic order of the server end of one embodiment of the application；
Fig. 2 is the flow diagram of the processing method of the phonetic order of the server end of another embodiment of the application；
Fig. 3 is the structural schematic diagram of the decision tree of one embodiment of the application；
Fig. 4 is the flow diagram of the processing method of the phonetic order of the server end of another embodiment of the application；
Fig. 5 is the flow diagram of the processing method of the phonetic order of the server end of another embodiment of the application；
Fig. 6 is the flow diagram of the processing method of the phonetic order of the client of one embodiment of the application；
Fig. 7 is the flow diagram of the processing method of the phonetic order of the client of another embodiment of the application；
Fig. 8 is the flow diagram of the processing method of the phonetic order of the client of another embodiment of the application；
Fig. 9 is the structural schematic diagram of the processing unit of the phonetic order of the server end of one embodiment of the application；
Figure 10 is the structural schematic diagram of the processing unit of the phonetic order of one embodiment of the application；
Figure 11 is the structural schematic diagram of the smart machine of one specific embodiment of the application；
Specific embodiment
The details for illustrating the application by embodiment with reference to the accompanying drawing is more advantageous in this way and understands that the application's is interior
Hold, but the application can by it is a variety of be different from specific embodiment in a manner of implement, those skilled in the art can without prejudice to
The prior art is combined to do similar popularization in the case where the application intension, therefore the application is not by the specific embodiment of following discloses
Limitation.
In this application, " first ", " second ", " third ", " the 4th " etc. are only used for mutual differentiation, rather than indicate important
Degree and sequence and each other existing premise etc..
In this application, processing method, device, smart machine and the storage medium of a kind of phonetic order are provided, under
It is described in detail one by one in the embodiment in face.
A kind of processing method of the phonetic order of server end is disclosed in one embodiment of the application, it is described referring to Fig. 1
Method includes：
Step 101：Receive the phonetic order for coming that self terminal includes user's original intent；
Step 102：Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；
Step 103：The text information is parsed, determines that parsing corresponding to the text information is intended to；
Step 104：It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent out
It send to the terminal；
Step 105：It determines that the original intent is unmet, the phonetic order, text information and parsing is intended to
Error sample is labeled as to save to error sample library.
The above method according to the reaction pattern of human-computer interaction determine whether obtain error sample, it is no longer necessary to user manually into
The operation of row error label, to increase a possibility that getting error sample.
Determining that the original intent is unmet in one embodiment according to the application, in step 105 includes：
The parsing for repeating to receive same user in preset period of time is intended to identical phonetic order；Or
It receives and is intended to the phonetic order, text information and parsing from the terminal to be labeled as the letter of error sample
Breath.
By taking the first situation as an example, it is assumed that user wants being named as of one well-known shop Pizza of purchase " dustbin "
Pizza, then it inputs phonetic order " taking out a dustbin Pizza " by mobile phone, and received server-side refers to the voice
There is mistake during speech recognition after order, phonetic order is converted into text information " one dustbin of mother
Pizza ", and the matched refuse collection of intention removal search that parsing makes mistake is held according to the text information of this mistake and services public affairs
Department.Since parsing is intended to be intended to not be inconsistent with user, it is unable to satisfy user's request, user is likely to attempt input voice again and refers to
" taking out a dustbin Pizza " is enabled, current server, which correctly changes phonetic order in speech recognition process, switchs to text envelope
" taking out a dustbin Pizza " is ceased, is gone out according to the Context resolution of text information and is correctly intended to, sale Pizza/ is searched and drapes over one's shoulders
Sa/Piza restaurant inventory simultaneously provides the map route in the shop of arriving.It is well-known that user has found that family in the restaurant inventory of offer
The information such as the address in the shop Pizza and phone, and successfully have subscribed the Pizza for being named as " dustbin ".In this process, quilt
Phonetic order " taking out a dustbin Pizza ", text information " one dustbin Pizza of mother " and the parsing meaning of wrong identification
Figure " searching for matched refuse collection service company " is noted as error sample and saves into error sample library, for improving mistake
The accuracy of pattern detection model.Reduce a possibility that wrong identification or parsing hereafter occurs.
It provides a kind of server end in the embodiment and determines the unmet judgment mode of the original intent, i.e., it is logical
Cross detect it is a certain represent it is identical parsing be intended to phonetic order whether repeat to transmit determination in a short time by same user be
No acquisition error sample, this method are omitted the tedious steps that user carries out error label feedback manually, obtain error sample
A possibility that greatly improve.
Second situation, if received server-side arrival self terminal anticipates the phonetic order, text information and parsing
Figure is labeled as the information of error sample, it is determined that the original intent of user is unmet, by the phonetic order, text information
Error sample is labeled as with parsing intention to save to error sample library.
In another embodiment according to the application, as shown in Fig. 2, wherein step 201 to 204 with as shown in Figure 1 side
Step 101 in method constructs the detection model that can learn, for examining to 104 identical in a manner of decision tree in step 205
The identical phonetic order of parsing intention for whether repeating to receive same user in preset period of time is surveyed, so that it is determined that described
Whether original intent is met, and conciliates the phonetic order, text information in the case where original intent is unmet
Analysis intention is labeled as error sample and saves to error sample library.
Decision tree is a kind of prediction model in machine learning, it indicates that one kind between object properties and object value is reflected
It penetrates, each of decision tree node indicates that the Rule of judgment of object properties, branching representation meet the object of node condition.Certainly
The leaf node of plan tree indicates prediction result belonging to object.
Fig. 3 shows the structural schematic diagram of the decision tree in a specific embodiment, for detecting whether in preset period of time
The parsing that interior repetition receives same user is intended to identical phonetic order, and then determines whether current phonetic order, text
This information and parsing intention are labeled as error sample and save to error sample library.
In the decision tree of this structure there are three the attributes of judgment basis：
The first, whether repeat to receive the instruction of identical intention；
The second, whether the instruction of the described identical intention is from same subscriber；
Third, whether be repeated 2 times in two minutes it is above.
Each of decision tree node indicates that the Rule of judgment of object properties, branching representation meet pair of node condition
As.Such as：Server repetition receives the phonetic order of identical intention, described instruction and repeats from same subscriber, in 2 minutes
3 times.Judge that the situation meets right branch (YES) by the root node of decision tree；Judge whether again from identical use
Family meets right branch (YES)；Then judge whether to be repeated 2 times in 2 minutes above, meet right branch (YES), work as cause
Condition is just fallen on the leaf node of " original intent is unmet ", and current phonetic order, text information and parsing are intended to
Error sample is labeled as to save to error sample library.
The building of the decision tree can using ID3 algorithm (Iterative Dichotomiser 3), C4.5 algorithm or
CART algorithm etc..
The ID3 algorithm is a kind of greedy algorithm, for constructing decision tree.ID3 algorithm originates from concept learning system, with
The decrease speed of comentropy is to choose the standard of testing attribute, i.e., has most what each node selection was not yet used to divide
The attribute of high information gain then proceedes to this process as the criteria for classifying, until the decision tree of generation can perfect classification based training
Sample.
The C4.5 algorithm and ID3 algorithm are solved using greedy algorithm, and the two is the difference is that classification
The foundation of decision is different.When carrying out categorised decision with information gain, it is partial to the more feature of value, C4.5 is namely based on letter
Cease the categorised decision method of the ratio of gains.Therefore, C4.5 algorithm is identical with ID3 in recurrence in structure, and difference is that choosing
Select information gain than maximum when depending on disconnected feature.
The CART algorithm is also known as post-class processing algorithm, and the post-class processing is binary tree, therefore CART algorithm
Dichotomy can simplify the scale of decision tree, improve the efficiency for generating decision tree.
The sharpest edges of decision Tree algorithms are to the self study of implementation model, it is only necessary to carry out to training example preferable
Mark, it will be able to train the good error sample detection model of effect.
In the case where the dimension relatively depth of decision tree, it is easy to there is the phenomenon that over-fitting, so-called over-fitting refer in order to
Unanimously assumed and makes to assume to become over stringent.Avoiding over-fitting is one of core missions of classifier design.In order to anti-
Other than the pruning method except through limiting decision tree dimension, it is random can also to construct a large amount of decision tree composition for only over-fitting
Forest prevents over-fitting, the disadvantage that avoids decision tree generalization ability weak.In other words, single decision tree is there may be over-fitting,
But over-fitting can be eliminated by the increase of range.Random forest technology can preferably handle high latitude data,
Training can be rapidly completed in the case where multiple features.In addition, random forest is able to detect that between feature in the training process
It influences each other to predict whether sample belongs to error sample.
Therefore, error sample can be filtered by the detection model that the method optimizing of random forest can learn.
It is a kind of it is typical building random forest method include：
N sample is randomly selected from sample set；
K attribute is randomly selected from all properties, is selected optimal segmentation attribute as node and is established decision tree；
It repeats two above step m times, establishes m decision tree；
This m decision tree forms random forest, the voting result by way of ballot, which kind of determination data belongs to.Institute
It states voting machine and is formed with that veto by one vote system, the minority is subordinate to the majority, weighted majority etc..
It, can be by being modeled to user's customary model used in everyday, from wrong sample by decision tree or random forest technology
Some special users, such as tester are rejected in this library, operate generated error sample.
Below by taking a concrete application as an example, the effect of random forest technology in this application is illustrated, wherein：
Class of subscriber is divided into：Tester and ordinary user.
The corresponding feature for classification of each decision tree in random forest, if total characteristic number is 3 forests
In be just corresponding with 3 decision trees, decision tree herein uses post-class processing.
First decision tree classified for feature " the daily average duration for using phonetic order " is shown in table 1
Parameter：
Second decision tree classified for feature " the daily par for uploading phonetic order " is shown in table 2
Parameter：
| Upload the quantity of voice | Tester | Ordinary user |
| It is daily to be more than or equal to 500 | 75% | 1% |
| It is daily to be more than or equal to 100 | 85% | 8% |
| It is daily to be less than or equal to 50 | 15% | 75% |
| It is daily to be less than or equal to 10 | 1% | 35% |
The ginseng for the third decision tree classified for feature " par of daily marking error " is shown in table 1
Number：
| Upload the quantity of voice | Tester | Ordinary user |
| Daily more than 100 | 80% | 2% |
| It is daily to be more than or equal to 50 | 92% | 15% |
| It is daily to be less than or equal to 20 | 30% | 55% |
| It is daily to be less than or equal to 10 | 1% | 30% |
According to the classification results of above-mentioned three decision trees, user's classification can be established for the information of some specific user
Distribution situation：
| Feature | Characteristic value | Tester | Ordinary user |
| The daily average duration for using phonetic order | 7 | 95% | 5% |
| The daily par for uploading phonetic order | 100 | 85% | 8% |
| The par of daily marking error | 50 | 92% | 15% |
It finally draws a conclusion, it is tester which, which has about 91% probability, and about 9% probability is ordinary user, institute
Finally to assert that the user belongs to tester, error sample caused by the user's operation is rejected from error sample library.
In another embodiment according to the application, the method includes：
Step 401：Receive the phonetic order for coming that self terminal includes user's original intent；
Step 402：Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；
Step 403：The text information is parsed, determines that parsing corresponding to the text information is intended to；
Step 404：It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent out
It send to the terminal, saves the record of resource retrieval corresponding with the parsing intention；
Step 405：It determines that the original intent is unmet, the phonetic order, text information, parsing is intended to mark
Note is that error sample is saved to error sample library；
Step 406：If determining that the parsing intention is unmet according to preset rules based on retrieval record
The reason is that then the error sample is rejected from the error sample library caused by resource retrieval.
Parsing caused by resource retrieval reason is intended to unmet situation and includes：
Network connection error leads to not obtain resource retrieval result；
The mode mistake of retrieval leads to the resource retrieval result of mistake；Or
Because the limitation of search library leads to not obtain the resource retrieval result needed.
It may be that speech recognition errors or semantic interpretation generate in the process since user is intended to unsatisfied situation both,
Be also likely to be as resource retrieval failure or mistake caused by, when carrying out error sample preservation, at the same save resource examine
The record of rope, can subsequently through modes such as artificial reinspections, by this kind of speech recognition and semantic parsing it is correct and merely because
User caused by retrieval failure or mistake is intended to unsatisfied interference error sample and rejects from error sample library, to improve
The accuracy of error sample detection model.
In a specific application, the phonetic order that server receives user's input " plays film《AABBCC》",
" film is played by the text information that speech recognition obtains the phonetic order《AABBCC》", after parsing text information, into
It is not found when row resource retrieval entitled《AABBCC》Film video resource, subsequent user repeatedly inputs phonetic order, but by
In retrieval less than matched film video resource, user is intended to be unable to get satisfaction always, therefore phonetic order " plays film
《AABBCC》", text information corresponding with the phonetic order and parsing be intended to and money corresponding with the parsing intention
The record of source retrieval is all noted as error sample and saves into error sample library.Obviously, it is parsed in speech recognition and semanteme
Do not occur mistake in link, therefore, which can be removed from error sample library by artificial screening.
In another embodiment according to the application, to avoid the cumbersome of artificial screening, money is saved in retrieval record
The matching degree of source and request interferes the automatic screening of wrong data sample to realize, the method includes：
Step 501：Receive the phonetic order for coming that self terminal includes user's original intent；
Step 502：Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；
Step 503：The text information is parsed, determines that parsing corresponding to the text information is intended to；
Step 504：It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent out
It send to the terminal, saves the record of resource retrieval corresponding with the parsing intention；
Step 505：The matching degree that the resource retrieval and the parsing are intended to is calculated, and by the matching degree being calculated
Value is stored in retrieval record；
Step 506：It determines that the original intent is unmet, the phonetic order, text information and parsing is intended to
Error sample is labeled as to save to error sample library；
Step 507：The value of the matching degree is less than preset threshold, then by the error sample from the error sample library
It rejects.
Through the above steps, the mistake for not being able to satisfy user's intention and mark can will be caused due to resource retrieval
Sample screening comes out, and is rejected from error sample library, and the accuracy of error sample detection model is further increased.
In an alternate embodiment of the invention, can by it is described cause not to be able to satisfy due to resource retrieval user be intended to and
The error sample of mark just filters this out before saving to error sample library.
There are many kinds of the modes for the matching degree that computing resource retrieval is intended to parsing, is illustrated below with a concrete application
Explanation.If the KTV that user wants to go to one entitled " lollipop ", which sings, wants relevant search information, user inputs voice and refers to thus
" lollipop KTV singing " is enabled, which is carried out speech recognition to have obtained corresponding text information being that " lollipop KTV is sung
Song ", obtains three keywords " lollipop ", " KTV " and " singing " according to the fractionation to text information, is solved according to keyword
Analysis is intended to the KTV in search name comprising keyword " lollipop ".But due to there is no one entitled " lollipop " KTV,
Therefore following 4 kinds of search results can only be provided as feedback：
1. providing the information for not including the place that can be sung of keyword " lollipop " in title；
2. phoning comprising keyword " lollipop " and/or " KTV " and/or the contact person of " singing "；
3. the schedule of " lollipop KTV singing " is added in calendar；
4. playing the song including keyword " lollipop " and/or " KTV ".
Obviously regardless of providing any original intent for being all unable to satisfy user in above-mentioned four kinds of feedbacks, but due to resource
Retrieval with it is described parsing be intended to matching degree it is not high, it is assumed that the calculated matching degree numerical value of the first situation be 50%, second
The calculated matching degree of situation be 40%, the calculated matching degree of the third situation be 30%, the 4th kind situation calculated
It is 15%, respectively less than preset threshold 70% with degree, even if this for receiving same user's transmission is repeated several times in server at this time
Same voice instruction, finally all can because of resource retrieval and it is described parsing be intended to matching degree value be less than preset threshold, and incite somebody to action
Corresponding error sample is rejected from the error sample library.
The calculating of matching degree depends on the type of resource retrieval, for retrieving a song.Phonetic order is to play song
Bent S1, S1 represent the character string of song title.But there is mistake in speech recognition or semantic parsing, cause finally to obtain
The song title for including in parsing intention is that character string S2, S1 and S2 respectively correspond pinyin character string P1 and P2, passes through formula M=
1-d/max (len (p1), len (p2)) calculates the value of the matching degree using character string S2 as search condition, and wherein M is matching degree,
D is the editing distance of P1 and P2, and len (P1) is the length of pinyin character string P1, and len (P2) is the length of pinyin character string P2,
Max (len (P1), len (P2)) takes the biggish numerical value of string length in the two.
In one embodiment of the application, a kind of processing method of the phonetic order of client is disclosed, it is described referring to Fig. 6
Method includes：
Step 601：The phonetic order is simultaneously sent to server by the phonetic order of original intent of the acquisition comprising user；
Step 602：Resource needed for executing the phonetic order is obtained from server end；
Step 603：It determines that the original intent is unmet, sends to the server by the phonetic order and base
The text information of speech recognition acquisition is carried out in the phonetic order and the parsing of parsing acquisition is carried out to the text information
It is intended to be labeled as the information of error sample.
The above method decides whether acquisition error sample according to the reaction pattern of human-computer interaction, carries out mistake manually without user
The accidentally operation of mark, increases a possibility that getting error sample.
Optionally, the processing side of the phonetic order of another client is provided in another embodiment according to the application
Method, referring to Fig. 7, the method includes：
Step 701：The phonetic order is simultaneously sent to server by the phonetic order of original intent of the acquisition comprising user；
Step 702：Resource needed for executing the phonetic order is obtained from server end；
Step 703：The information of the corresponding execution movement of the phonetic order is provided based on the resource got；
Step 704：The instruction for abandoning executing the corresponding execution movement of the phonetic order is captured, determines the original meaning
Scheme it is unmet, to the server send by the phonetic order and based on the phonetic order carry out speech recognition acquisition
Text information and parsing acquisition is carried out to the text information parsing be intended to be labeled as the information of error sample.
Feedback information is provided a user based on the resource got in the above method, is informed by the feedback information
The user client subsequent action to be carried out.The feedback information can be text feedback and be also possible to through TTS (text turn
Voice Text To Speech) technology realize voice feedback, text can be realized by common text-to-speech converting unit
Conversion of the information to audio-frequency information.User sees or hears the feedback information, it will be able to judge that can request be met.
In general, user only be intended to be not being met in the case where, the execution of subsequent action just can be actively abandoned, so if client
Termination receives the instruction that user abandons executing subsequent action, can estimate needed for the execution phonetic order that client is got
The resource wanted is not able to satisfy the actual request of user, carries out voice by the phonetic order and based on the phonetic order at this time
Identify that the text information obtained and the parsing intention for carrying out parsing acquisition to the text information are labeled as error sample and save
To error sample library, the complicated processes of the manual marking error sample of user are not only omitted, but also are complying fully with the normal of user
Automatic collection error sample in the state of rule operating habit, which greatly enhances the probability for getting effective error sample.
A kind of processing method of the phonetic order of client is provided in another embodiment according to the application, referring to
Fig. 8, the method includes：
Step 801：The phonetic order is simultaneously sent to server by the phonetic order of original intent of the acquisition comprising user；
Step 802：Resource needed for executing the phonetic order is obtained from server end；
Step 803：Execute that the phonetic order is corresponding to execute movement based on the resource that gets；
Step 804：The instruction that the corresponding execution movement of the phonetic order is terminated in scheduled time threshold value is captured, really
The fixed original intent is unmet, sends to the server by the phonetic order and is carried out based on the phonetic order
The text information of speech recognition acquisition and the parsing intention for carrying out parsing acquisition to the text information are labeled as error sample
Information.
For example, the phonetic order of acquisition user's input " plays video《ABCC》", and the phonetic order is uploaded into clothes
Business device end；Mistake occurs when carrying out speech recognition by server end, the text information that phonetic order is identified as mistake " is played and regarded
Frequently《ADCC》", it is intended to play according to the parsing that the text information of mistake parses entitled《ADCC》Video, and according to final
Determining parsing is intended to retrieval and executes video resource required for the parsing is intended to《ADCC》；Client is got《ADCC》's
Start to play after video resource, user has found that the video played is not that its request plays at this time《ABCC》Therefore it is broadcast in video
It puts to have issued for 5 seconds and terminates the instruction of broadcasting, client captures this instruction of user, thereby determines that user's is original
Be intended to it is unmet, by the phonetic order " play video《ABCC》" and based on the phonetic order carry out speech recognition obtain
The text information obtained " plays video《ADCC》" and to the text information carry out parsing acquisition parsing be intended to " play it is entitled
《ADCC》Video " be labeled as error sample and save to error sample library.
Also the mode that a kind of routine operation habit for meeting user is had chosen in the above method realizes oneself of error sample
Dynamic acquisition, is omitted the complicated processes of the manual marking error sample of user, improves the probability for getting effective error sample.
One embodiment of the application discloses a kind of processing unit of the phonetic order of server end, referring to Fig. 9, described device
Including：Receiving module 901, speech recognition module 902, parsing module 903, resource retrieval module 904, the detection of the first error sample
Module 905 and error sample library 906；Wherein, the receiving module 901 is configured as reception to carry out self terminal including the original meaning of user
The phonetic order of figure；The speech recognition module 902 is configured as carrying out speech recognition to the phonetic order, generates institute's predicate
The text information of sound instruction；The parsing module 903 is configured as parsing the text information, determines the text envelope
The corresponding parsing of breath is intended to；The resource retrieval module 904 is configured as being intended to retrieval execution institute's predicate according to the parsing
Resource needed for sound instruction, and the resource is sent to the terminal；The first error sample detection module 905 is configured
It is unmet for the determination original intent, it is intended to the phonetic order, text information and parsing to be labeled as error sample
It saves to error sample library 906；The error sample library 906 is configured as storing the error sample.
It can be decided whether to obtain error sample according to the reaction pattern of human-computer interaction by above-mentioned apparatus, be not necessarily to user hand
The dynamic operation for carrying out error label, increases a possibility that getting error sample.
In another embodiment according to the application, the first error sample detection module 905 determines the original meaning
Scheme unmet include：
The parsing for repeating to receive same user in preset period of time is intended to identical phonetic order；Or
It receives and is intended to the phonetic order, text information and parsing from the terminal to be labeled as the letter of error sample
Breath.
The detection module 905 of the device repeats the identical parsing meaning of the representative uploaded by capturing same user in a short time
The phonetic order of figure acquires the opportunity of error sample to determine, the cumbersome step that user carries out error label feedback manually is omitted
Suddenly, a possibility that obtaining error sample, greatly improves.Another judgment mode is to determine the original for that will meet client
The error sample for beginning to be intended to unmet judgment criteria is saved in error sample library.
In another embodiment according to the application, the first error sample detection module is configured as with decision tree
Mode constructs the detection model that can learn, for detecting whether repeating to receive the parsing of same user in preset period of time
It is intended to identical phonetic order, so that it is determined that whether the original intent is met, in the unmet feelings of original intent
The phonetic order, text information and parsing intention are labeled as error sample and saved to error sample library under condition.
The sharpest edges of decision Tree algorithms are to the self study of implementation model, it is only necessary to carry out to training example preferable
Mark, can error sample be able to be filtered by the detection model that the method optimizing of random forest can learn.
Random forests algorithm can prevent over-fitting, solve the weak disadvantage of decision tree generalization ability.
In another embodiment according to the application, the first error sample detection module 905 is additionally configured to：
Save the record of resource retrieval corresponding with the parsing intention；
If determining that the parsing is intended to unmet reason according to preset rules based on the retrieval record is money
Caused by the retrieval of source, then the error sample is rejected from the error sample library.
As user be intended to unsatisfied situation be also likely to be as caused by resource retrieval failure or mistake,
Device configured with the first error sample detection module 905 as described above saves money when carrying out error sample preservation
The record of source retrieval, can subsequently through modes such as artificial reinspections, by this kind of speech recognition and semantic parsing it is correct and only
It is rejected from error sample library because user caused by retrieval failure or mistake is intended to unsatisfied interference error sample, thus
Improve the accuracy of error sample detection model.
According to another embodiment of the application, the first error sample detection module 905 is additionally configured to：
Save the record of resource retrieval corresponding with the parsing intention；
The matching degree that the resource retrieval and the parsing are intended to is calculated, and the value for the matching degree being calculated is stored in
In retrieval record；
The value of the matching degree is less than preset threshold, then rejects the error sample from the error sample library.
Device configured with the first error sample detection module 905 as described above can will be due to resource retrieval
Cause the error sample screening for not being able to satisfy user's intention and marking to come out, and rejected from error sample library, further
Improve the accuracy of error sample detection model.
A kind of processing unit of the phonetic order of client, such as Figure 10 are disclosed in one embodiment according to the application
Shown, described device includes：Acquisition module 1001, execution module 1002 and the second error sample detection module 1003；It is described to adopt
Collection module 1001 is configured as the phonetic order of original intent of the acquisition comprising user and the phonetic order is sent to service
Device；The execution module 1002 is configured as resource needed for obtaining the execution phonetic order from server end；Described second
Error sample detection module 1003 is configured to determine that the original intent is unmet, and sending to the server will be described
Phonetic order and based on the phonetic order carry out speech recognition acquisition text information and the text information is solved
The parsing that analysis obtains is intended to be labeled as the information of error sample.
In order to facilitate the principle of description client and server cooperating, also shown in Figure 10 and client
The structure composition of the processing unit of the phonetic order of the server end of the processing unit cooperating of phonetic order.Wherein, described
Receiving module 1004 is configured as receiving the phonetic order comprising user's original intent that acquisition module 1001 described in terminal is sent；
The phonetic order is transmitted to speech recognition module 1005 and carries out speech recognition, generates the text information of the phonetic order；
The text information is transmitted to parsing module 1006 and determines that parsing corresponding to the text information is intended to；Later, resource is examined
Rope module 1007 is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent to institute
State the execution module 1002 of terminal；The execution module 1002 of terminal obtains needed for executing the phonetic order from server end
Resource after by the second error sample detection module 1003 determine user original intent whether met, however, it is determined that
The original intent is unmet then to be sent the phonetic order to server and carries out voice based on the phonetic order
Identify that the text information obtained and the parsing for carrying out parsing acquisition to the text information are intended to be labeled as the letter of error sample
Breath.Wherein the first error sample detection module 1003 can save the note of resource retrieval corresponding with the parsing intention
Record；If determining that the parsing is intended to unmet reason according to preset rules based on the retrieval record is resource retrieval
It is caused, then the error sample is rejected from the error sample library 1009.
The error sample library 1009 is generally located on server end, is also not precluded within client certainly for specific user
The possibility of specific faults sample database is set.
Above-mentioned apparatus can be decided whether according to the reaction pattern of human-computer interaction obtain error sample, without user manually into
The operation of row error label, increases a possibility that getting error sample.
According to one embodiment of the application, the execution module 1002 is additionally configured to refer in the acquisition execution voice
After resource required for enabling, the information of the corresponding execution movement of the phonetic order is provided based on the resource got；
The determination original intent is unmet to include：
Capture the instruction for abandoning executing the corresponding execution movement of the phonetic order.
The complicated processes of the manual marking error sample of user are not only omitted in the device for being configured with above-mentioned execution module 1002,
And the automatic collection error sample in the state that routine operation for complying fully with user is accustomed to, which greatly enhances got
Imitate the probability of error sample.
According to another embodiment of the application, the execution module 1002 is additionally configured to obtaining the execution voice
It required for instructing after resource, is based on, executes that the phonetic order is corresponding to execute movement based on the resource got；
The determination original intent is unmet to include：
Capture the instruction that the corresponding execution movement of the phonetic order is terminated in scheduled time threshold value.
The device for being configured with above-mentioned execution module 1002 also can be with a kind of side of routine operation habit for meeting user
Formula realizes the automatic collection of error sample, and the complicated processes of the manual marking error sample of user are omitted, improves and has got
The probability of the error sample of effect.
A kind of smart machine 1100 as shown in figure 11 is provided in one embodiment according to the application, including but not
It is limited to memory 1101, processor 1102 and is stored in the computer that can be run on memory 1101 and on processor 1102 to refer to
It enables, the processor 1102 realizes the processing method of foregoing phonetic order when executing described instruction.
A kind of exemplary scheme of above-mentioned smart machine for the present embodiment.It should be noted that the skill of the smart machine
Art scheme and the processing method of phonetic order above-mentioned belong to same design, and the technical solution of the smart machine is not described in detail
Detail content, may refer to the description of the technical solution of the processing method of above-mentioned phonetic order.
A kind of computer readable storage medium is provided in one embodiment according to the application, is stored thereon with calculating
Machine instruction, realizes the processing method for weighing foregoing phonetic order when described instruction is executed by processor.
The computer instruction includes computer program code, the computer program code can for source code form,
Object identification code form, executable file or certain intermediate forms etc..The computer-readable medium may include：Institute can be carried
State any entity or device, recording medium, USB flash disk, mobile hard disk, magnetic disk, CD, the computer storage of computer program code
Device, 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 should be noted that the computer-readable medium include it is interior
Increase and decrease appropriate can be carried out according to the requirement made laws in jurisdiction with patent practice by holding, such as in certain jurisdictions of courts
Area does not include electric carrier signal and telecommunication signal according to legislation and patent practice, computer-readable medium.
A kind of exemplary scheme of above-mentioned computer readable storage medium for the present embodiment.It should be noted that this is deposited
The technical solution of storage media and the processing method of phonetic order above-mentioned belong to same design, and the technical solution of storage medium is not detailed
The detail content carefully described may refer to the description of the technical solution of the processing method of above-mentioned phonetic order.
It should be noted that for the various method embodiments described above, describing for simplicity, therefore, it is stated as a series of
Combination of actions, but those skilled in the art should understand that, the application is not limited by the described action sequence because
According to the application, certain steps can use other sequences or carry out simultaneously.Secondly, those skilled in the art should also know
It knows, the embodiments described in the specification are all preferred embodiments, and related actions and modules might not all be this Shen
It please be necessary.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiments.
The application preferred embodiment disclosed above is only intended to help to illustrate the application.There is no detailed for alternative embodiment
All details are described, also not limiting this application is only the specific embodiment.Obviously, according to the content of this specification,
It can make many modifications and variations.These embodiments are chosen and specifically described to this specification, is in order to preferably explain the application
Principle and practical application, so that skilled artisan be enable to better understand and utilize the application.The application is only
It is limited by claims and its full scope and equivalent.
Claims (10)
1. a kind of processing method of phonetic order, which is characterized in that the method includes：
Receive the phonetic order for coming that self terminal includes user's original intent；
Speech recognition is carried out to the phonetic order, generates the text information of the phonetic order；
The text information is parsed, determines that parsing corresponding to the text information is intended to；
It is intended to resource needed for retrieval executes the phonetic order according to the parsing, and the resource is sent to the end
End；
It determines that the original intent is unmet, is intended to the phonetic order, text information and parsing to be labeled as wrong sample
This preservation is to error sample library.
2. the method according to claim 1, wherein the determination original intent is unmet includes：
The parsing for repeating to receive same user in preset period of time is intended to identical phonetic order；Or it receives and comes from
The phonetic order, text information and parsing are intended to be labeled as the information of error sample by the terminal.
3. according to the method described in claim 2, it is characterized in that, being determined whether in a manner of decision tree in preset period of time
The parsing that interior repetition receives same user is intended to identical phonetic order.
4. method according to claim 1 or 2, which is characterized in that the method also includes：
Save the record of resource retrieval corresponding with the parsing intention；
If determining that the parsing is intended to unmet reason according to preset rules based on the retrieval record is resource inspection
Caused by rope, then the error sample is rejected from the error sample library.
5. method according to claim 1 or 2, which is characterized in that the method also includes：
Save the record of resource retrieval corresponding with the parsing intention；
The matching degree that the resource retrieval and the parsing are intended to is calculated, and the value for the matching degree being calculated is stored in retrieval
In record；
The value of the matching degree is less than preset threshold, then rejects the error sample from the error sample library.
6. a kind of processing method of phonetic order, which is characterized in that the method includes：
The phonetic order is simultaneously sent to server by the phonetic order of original intent of the acquisition comprising user；
Resource needed for executing the phonetic order is obtained from server end；
It determines that the original intent is unmet, send to the server by the phonetic order and is referred to based on the voice
The text information for carrying out speech recognition acquisition and the parsing intention for carrying out parsing acquisition to the text information is enabled to be labeled as mistake
The accidentally information of sample.
7. according to the method described in claim 6, it is characterized in that, obtaining resource required for executing the phonetic order
Afterwards, the method also includes：
The information of the corresponding execution movement of the phonetic order is provided based on the resource got；
The determination original intent is unmet to include：
Capture the instruction for abandoning executing the corresponding execution movement of the phonetic order.
8. according to the method described in claim 6, it is characterized in that, after obtaining resource required for executing the phonetic order,
The method also includes：
Execute that the phonetic order is corresponding to execute movement based on the resource that gets；
The determination original intent is unmet to include：
Capture the instruction that the corresponding execution movement of the phonetic order is terminated in scheduled time threshold value.
9. a kind of processing unit of phonetic order, which is characterized in that described device includes：Acquisition module, execution module and second
Error sample detection module；The acquisition module is configured as the phonetic order of original intent of the acquisition comprising user and will be described
Phonetic order is sent to server；The execution module is configured as obtaining needed for executing the phonetic order from server end
Resource；The second error sample detection module is configured to determine that the original intent is unmet, to the server
Send by the phonetic order and based on the phonetic order carry out speech recognition acquisition text information and to the text
The parsing that information carries out parsing acquisition is intended to be labeled as the information of error sample.
10. a kind of computer readable storage medium, is stored thereon with computer instruction, which is characterized in that the instruction is by processor
The processing method of phonetic order described in any one of claim 1 to 5 or 6 to 8 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810233853.8A CN108932945B (en) | 2018-03-21 | 2018-03-21 | Voice instruction processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810233853.8A CN108932945B (en) | 2018-03-21 | 2018-03-21 | Voice instruction processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108932945A true CN108932945A (en) | 2018-12-04 |
| CN108932945B CN108932945B (en) | 2021-08-31 |
Family
ID=64449316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810233853.8A Active CN108932945B (en) | 2018-03-21 | 2018-03-21 | Voice instruction processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108932945B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110010128A (en) * | 2019-04-09 | 2019-07-12 | 天津松下汽车电子开发有限公司 | A kind of sound control method and system of high discrimination |
| CN110706707A (en) * | 2019-11-13 | 2020-01-17 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device and computer-readable storage medium for voice interaction |
| CN110838284A (en) * | 2019-11-19 | 2020-02-25 | 大众问问(北京)信息科技有限公司 | Method and device for processing voice recognition result and computer equipment |
| CN110942772A (en) * | 2019-11-21 | 2020-03-31 | 新华三大数据技术有限公司 | Voice sample collection method and device |
| CN111586469A (en) * | 2020-05-12 | 2020-08-25 | 腾讯科技（深圳）有限公司 | Bullet screen display method and device and electronic equipment |
| CN111741354A (en) * | 2020-06-01 | 2020-10-02 | 深圳康佳电子科技有限公司 | Method, system and storage medium for assisting voice interaction based on interface elements |
| WO2020199963A1 (en) * | 2019-03-29 | 2020-10-08 | 时时同云科技(成都)有限责任公司 | Method for food ordering robot to identify food ordering intention of user, and robot |
| CN111914078A (en) * | 2020-08-13 | 2020-11-10 | 北京捷通华声科技股份有限公司 | Data processing method and device |
| CN112632962A (en) * | 2020-05-20 | 2021-04-09 | 华为技术有限公司 | Method and device for realizing natural language understanding in human-computer interaction system |
| CN113791545A (en) * | 2020-07-10 | 2021-12-14 | 北京沃东天骏信息技术有限公司 | Smart home equipment testing method and device, electronic equipment and readable storage medium |
| CN113889108A (en) * | 2021-10-19 | 2022-01-04 | 清华大学 | Error correction method and system for voice operation |
| CN114360515A (en) * | 2021-12-09 | 2022-04-15 | 北京声智科技有限公司 | Information processing method, information processing apparatus, electronic device, information processing medium, and computer program product |
| CN115762513A (en) * | 2022-11-03 | 2023-03-07 | 深圳市品声科技有限公司 | Voice-controlled split wireless audio communication method and system |
| WO2023207327A1 (en) * | 2022-04-26 | 2023-11-02 | 青岛海尔洗衣机有限公司 | Speech recognition accuracy testing method and apparatus, and device and storage medium |
| CN113889108B (en) * | 2021-10-19 | 2024-06-25 | 清华大学 | Error correction method and system for voice operation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1471502A1 (en) * | 2003-04-25 | 2004-10-27 | Sony International (Europe) GmbH | Method for correcting a text produced by speech recognition |
| CN103198828A (en) * | 2013-04-03 | 2013-07-10 | 中金数据***有限公司 | Method and system of construction of voice corpus |
| CN104252864A (en) * | 2013-06-28 | 2014-12-31 | 国际商业机器公司 | Real-time speech analysis method and system |
| CN107146607A (en) * | 2017-04-10 | 2017-09-08 | 北京猎户星空科技有限公司 | Modification method, the apparatus and system of smart machine interactive information |
| JP6232282B2 (en) * | 2013-12-26 | 2017-11-15 | 日本放送協会 | Speech recognition error correction device |
-
2018
- 2018-03-21 CN CN201810233853.8A patent/CN108932945B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1471502A1 (en) * | 2003-04-25 | 2004-10-27 | Sony International (Europe) GmbH | Method for correcting a text produced by speech recognition |
| CN103198828A (en) * | 2013-04-03 | 2013-07-10 | 中金数据***有限公司 | Method and system of construction of voice corpus |
| CN104252864A (en) * | 2013-06-28 | 2014-12-31 | 国际商业机器公司 | Real-time speech analysis method and system |
| JP6232282B2 (en) * | 2013-12-26 | 2017-11-15 | 日本放送協会 | Speech recognition error correction device |
| CN107146607A (en) * | 2017-04-10 | 2017-09-08 | 北京猎户星空科技有限公司 | Modification method, the apparatus and system of smart machine interactive information |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020199963A1 (en) * | 2019-03-29 | 2020-10-08 | 时时同云科技(成都)有限责任公司 | Method for food ordering robot to identify food ordering intention of user, and robot |
| CN110010128A (en) * | 2019-04-09 | 2019-07-12 | 天津松下汽车电子开发有限公司 | A kind of sound control method and system of high discrimination |
| CN110706707A (en) * | 2019-11-13 | 2020-01-17 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device and computer-readable storage medium for voice interaction |
| US11393490B2 (en) | 2019-11-13 | 2022-07-19 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method, apparatus, device and computer-readable storage medium for voice interaction |
| CN110838284B (en) * | 2019-11-19 | 2022-06-14 | 大众问问(北京)信息科技有限公司 | Method and device for processing voice recognition result and computer equipment |
| CN110838284A (en) * | 2019-11-19 | 2020-02-25 | 大众问问(北京)信息科技有限公司 | Method and device for processing voice recognition result and computer equipment |
| CN110942772A (en) * | 2019-11-21 | 2020-03-31 | 新华三大数据技术有限公司 | Voice sample collection method and device |
| CN111586469A (en) * | 2020-05-12 | 2020-08-25 | 腾讯科技（深圳）有限公司 | Bullet screen display method and device and electronic equipment |
| CN111586469B (en) * | 2020-05-12 | 2021-10-26 | 腾讯科技（深圳）有限公司 | Bullet screen display method and device and electronic equipment |
| CN112632962B (en) * | 2020-05-20 | 2023-11-17 | 华为技术有限公司 | Method and device for realizing natural language understanding in man-machine interaction system |
| CN112632962A (en) * | 2020-05-20 | 2021-04-09 | 华为技术有限公司 | Method and device for realizing natural language understanding in human-computer interaction system |
| CN111741354B (en) * | 2020-06-01 | 2022-10-04 | 深圳康佳电子科技有限公司 | Method, system and storage medium for assisting voice interaction based on interface elements |
| CN111741354A (en) * | 2020-06-01 | 2020-10-02 | 深圳康佳电子科技有限公司 | Method, system and storage medium for assisting voice interaction based on interface elements |
| CN113791545A (en) * | 2020-07-10 | 2021-12-14 | 北京沃东天骏信息技术有限公司 | Smart home equipment testing method and device, electronic equipment and readable storage medium |
| CN111914078A (en) * | 2020-08-13 | 2020-11-10 | 北京捷通华声科技股份有限公司 | Data processing method and device |
| CN113889108A (en) * | 2021-10-19 | 2022-01-04 | 清华大学 | Error correction method and system for voice operation |
| CN113889108B (en) * | 2021-10-19 | 2024-06-25 | 清华大学 | Error correction method and system for voice operation |
| CN114360515A (en) * | 2021-12-09 | 2022-04-15 | 北京声智科技有限公司 | Information processing method, information processing apparatus, electronic device, information processing medium, and computer program product |
| WO2023207327A1 (en) * | 2022-04-26 | 2023-11-02 | 青岛海尔洗衣机有限公司 | Speech recognition accuracy testing method and apparatus, and device and storage medium |
| CN115762513A (en) * | 2022-11-03 | 2023-03-07 | 深圳市品声科技有限公司 | Voice-controlled split wireless audio communication method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108932945B (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108932945A (en) | A kind of processing method and processing device of phonetic order | |
| CN108345690B (en) | Intelligent question and answer method and system | |
| CN107329967B (en) | Question answering system and method based on deep learning | |
| CN107797984B (en) | Intelligent interaction method, equipment and storage medium | |
| CN111182162B (en) | Telephone quality inspection method, device, equipment and storage medium based on artificial intelligence | |
| CN108197282B (en) | File data classification method and device, terminal, server and storage medium | |
| CN111210842B (en) | Voice quality inspection method, device, terminal and computer readable storage medium | |
| CN108304372A (en) | Entity extraction method and apparatus, computer equipment and storage medium | |
| CN109360550A (en) | Test method, device, equipment and the storage medium of voice interactive system | |
| CN110364146A (en) | Audio recognition method, device, speech recognition apparatus and storage medium | |
| WO2017183242A1 (en) | Information processing device and information processing method | |
| CN106682137A (en) | Intelligent stock investment adviser questioning-answering interaction method and intelligent stock investment adviser questioning-answering interaction system | |
| CN106407393A (en) | An information processing method and device for intelligent apparatuses | |
| CN111178081B (en) | Semantic recognition method, server, electronic device and computer storage medium | |
| CN111460145A (en) | Learning resource recommendation method, device and storage medium | |
| CN112364622B (en) | Dialogue text analysis method, device, electronic device and storage medium | |
| CN111753021A (en) | Method, device and equipment for constructing knowledge graph and readable storage medium | |
| CN108710653B (en) | On-demand method, device and system for reading book | |
| CN103294670A (en) | Searching method and system based on word list | |
| CN111678531B (en) | Subway path planning method based on LightGBM | |
| CN115952770B (en) | Data standardization processing method and device, electronic equipment and storage medium | |
| CN110047473B (en) | Man-machine cooperative interaction method and system | |
| CN116306974A (en) | Model training method and device of question-answering system, electronic equipment and storage medium | |
| CN110858234A (en) | Method and device for pushing information according to human emotion | |
| CN115836288A (en) | Method and apparatus for generating training data |
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 |