CN106875940A - A kind of Machine self-learning based on neutral net builds knowledge mapping training method - Google Patents
A kind of Machine self-learning based on neutral net builds knowledge mapping training method Download PDFInfo
- Publication number
- CN106875940A CN106875940A CN201710127387.0A CN201710127387A CN106875940A CN 106875940 A CN106875940 A CN 106875940A CN 201710127387 A CN201710127387 A CN 201710127387A CN 106875940 A CN106875940 A CN 106875940A
- Authority
- CN
- China
- Prior art keywords
- sentence
- answer
- vector
- lambda
- word
- 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
- 238000012549 training Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000007935 neutral effect Effects 0.000 title claims abstract description 25
- 238000013507 mapping Methods 0.000 title claims abstract description 19
- 239000013598 vector Substances 0.000 claims abstract description 53
- 210000005036 nerve Anatomy 0.000 claims abstract description 24
- 238000001228 spectrum Methods 0.000 claims description 25
- 230000003595 spectral effect Effects 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 238000001514 detection method Methods 0.000 claims description 4
- 230000001537 neural effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006854 communication Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing 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
- G10L15/08—Speech classification or search
- G10L15/18—Speech classification or search using natural language modelling
- G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
-
- 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
- G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
- G10L15/063—Training
-
- 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
- G10L17/00—Speaker identification or verification techniques
- G10L17/22—Interactive procedures; Man-machine interfaces
-
- 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
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
- G10L21/0208—Noise filtering
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
- G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
-
- 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
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/72—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for transmitting results of analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Evolutionary Computation (AREA)
- Machine Translation (AREA)
Abstract
Knowledge mapping training method is built the invention discloses a kind of Machine self-learning based on neutral net, including：The sentence based on Nature condition that user sends is obtained, noise reduction is filtered to read statement using voice de-noising algorithm, it is determined that the feedback sentence of matching；If it does not exist, then according to neutral net dialog model, providing the answer of the sentence sent for user；Including：The coding layer that the user sends statement model is configured to first nerves network, and sending sentence to user in the first nerves network parses, and obtains for representing that user sends the first intermediate vector of statement semantics；The decoding layer of the dialogue generation model is configured to nervus opticus network, the intermediate vector is parsed in the nervus opticus network, obtain the vector group for representing sentence answer, the present invention is obtained in that less mean square error using threshold speech noise reduction algorithm, improves the signal to noise ratio of reconstructed speech signal.
Description
Technical field
Knowledge is built the present invention relates to field in intelligent robotics, more particularly to a kind of Machine self-learning based on neutral net
Collection of illustrative plates training method.
Background technology
Chat robots (chatterbot) are a programs for simulating human conversation or chat.Chat robots are produced
Raw the reason for is that developer is put into oneself answer interested in database, when a problem is thrown to chat robots,
It finds most close problem by similarity mode algorithm from database, then according to the corresponding relation of question and answer,
Most proper answer is provided, reply to it chats companion.
However, in the scene of current robot chat, when can not be found and user's request in robot knowledge base
Problem match same or like problem when, robot cannot to user return correctly suitably answer in other words
Case.
The limitation of prior art this respect also results in semantic reason in addition to the problem for causing robot knowledge base limited
The mistake of solution, so that user is poor with the experience of robot communication process.Additionally, in knowledge reasoning prior art knowledge
Reasoning process also has a certain limitation, and traditional knowledge reasoning is all to write some rules by application developer to push away solving knowledge
Reason problem.But, be exhaustive and formulate these rules and cannot imagine for developer.Because in natural language processing neck
Have forever in domain and can not write not complete rule.At this moment being accomplished by robot has the learning ability of oneself, and makes inferences.
The content of the invention
The present invention has designed and developed a kind of Machine self-learning based on neutral net and has built knowledge mapping training method, uses
Threshold speech noise reduction algorithm is obtained in that less mean square error, improves the signal to noise ratio of reconstructed speech signal.
A further object of the invention is to generate dialog model using neural metwork training, the model obtained using training,
Robot can be freely with user talk.
The present invention provide technical scheme be：
A kind of Machine self-learning based on neutral net builds knowledge mapping training method, including：
The sentence based on Nature condition that user sends is obtained, read statement is filtered using threshold speech noise reduction algorithm
Ripple noise reduction, and obtain the classification of the sentence, and obtain the sentence above of the sentence, and the sentence above classification；
The sentence classification according to sentence, it is determined that the feedback sentence of matching；
If it does not exist, then according to neutral net dialog model, providing the answer of the sentence sent for user；Including：
The coding layer that the user sends statement model is configured to first nerves network, right in the first nerves network
User sends sentence and is parsed, and obtains for representing that user sends the first intermediate vector of statement semantics；
The decoding layer of the dialogue generation model is configured to nervus opticus network, to described in the nervus opticus network
Intermediate vector is parsed, and obtains the vector group for representing sentence answer；And
The vector group for representing sentence answer is exported as problem answers.
Preferably, it is when being parsed to the sentence that the user sends in the first nerves network including following
Step：
The sentence of user input is split into semantic minimum word unit in coding layer, obtains multiple word units, and
Obtain the attribute of each word unit respectively, select comprising at least one word more than information content as centre word, and by its with to
The form of amount is input to the input layer of the first nerves network as problem vector group；
In the output and previous moment of the hidden layer to the input layer of the first nerves network of the first nerves network
The output of the hidden layer of the first nerves network carries out semantic parsing, and carries out linear weighted combination, forms representative sentences justice
Intermediate vector.
Preferably, when being parsed to the intermediate vector in the nervus opticus network, comprise the following steps：
The intermediate vector is received in decoding layer, and the intermediate vector is defeated as the input layer of nervus opticus network
Enter；
The nervus opticus network hidden layer to described in the intermediate vector and previous moment from input layer
The output of the hidden layer of two neutral nets carries out semantic parsing, sequentially generates some single vectors, to form answer vector group, its
Described in the semanteme of each single vector in answer vector group correspond to the semanteme of minimum word unit in answer output statement；
The answer vector group is exported in the output layer of the nervus opticus network.
Preferably, after the answer vector group is exported as answer output statement, by answer output
Sentence is accordingly saved in knowledge base with dialogue read statement, is updated and is expanded with to knowledge base.
Preferably, after knowledge base matching primitives are carried out, according in knowledge base whether there is and it is described dialogue input language
The conversation sentence that the matching degree of sentence reaches predetermined value asks standard signal position to set, and according to the effective of request flag signal position
Property come decide whether request dialogue generation model provide answer.
Preferably, the linear weighted combination, comprises the following steps：
Step one：N groups center word data group is extracted in counting user read statement, wherein n is positive integer, in every group
The probability x that heart term data ZuλTian Nei centers word occursi, previous sentence center word data group λ Tian Nei centers word appearance
Probability yi, single argument regression model is set up,
yi=ω 'i·xi
Wherein, i is integer, i=1,2,3...... λ, ω 'iIt is the weighted regression coefficient in λ days；
Step 2：Solved using the least square method formula a kind of to step, the recurrence in λ days is calculated respectively
Coefficient estimate：
Wherein,It is the estimate of regression coefficient；xijFor jZu centers word data group ZhongiTian centers word occurs
Probability；It is the average value of j groups center word data group probability；yijIn in jth group user input sentence previous moment sentence
I-th day in heart term data group；The probability that center word occurs；It is putting down for j group previous moments center word data group probability
Average
Step 3：Normalized, the weighted value after being weighted：
Wherein, ωiIt is the weighted value after the weighting of user input sentence.
Preferably, the output collocation person of needing to use is selected and is used, and when exporting answer and being accurate, storage is being known
Know storehouse.
Preferably, the voice de-noising algorithm, including：
A, mute frame and speech frame are divided into by end-point detection by speech frame；
B, for mute frame, calculates the power spectral value of present frame as noise power spectrum estimate, for speech frame, calculates
Voice noise power Spectral Estimation value；
C, noise power spectrum estimate is subtracted by the power spectrum of speech frame, obtains the spectrum of the phonetic speech power after noise reduction；
D, the speech frame after noise reduction is drawn according to the phonetic speech power spectrum after noise reduction.
9th, the Machine self-learning based on neutral net according to claim 8 builds knowledge mapping training method, its
It is characterised by, the voice noise power Spectral Estimation value computing formula is：
Wherein, I is noise power spectrum energy；Threshold valueN believes for noise
Number frame number；J=1-5It is conversion coefficient, e is natural constant；π is pi；fcIt is the frequency of noise signal；τ (t)=
0.03t2+0.6t+0.1；T is decomposition scale, 1≤t≤4.
Beneficial effects of the present invention
The present invention has designed and developed a kind of Machine self-learning based on neutral net and has built knowledge mapping training method, uses
Threshold speech noise reduction algorithm is obtained in that less mean square error, improves the signal to noise ratio of reconstructed speech signal.
A further object of the invention is to generate dialog model using neural metwork training, the model obtained using training,
Robot can be freely with user talk.
Brief description of the drawings
Fig. 1 is the flow that the Machine self-learning based on neutral net of the present invention builds knowledge mapping training method
Figure.
Specific embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings, to make those skilled in the art with reference to specification text
Word can be implemented according to this.
As shown in figure 1, the Machine self-learning based on neutral net that the present invention is provided builds knowledge mapping training method, bag
Include：
S100：The sentence based on Nature condition that user sends is obtained, using threshold speech noise reduction algorithm to read statement
Be filtered noise reduction, and obtain the classification of the sentence, and obtain the sentence above of the sentence, and the sentence above class
Not；
S200：The sentence classification according to sentence, it is determined that the feedback sentence of matching；
S300：If it does not exist, then according to neutral net dialog model, providing the answer of the sentence sent for user；
Including：
S310：User send statement model coding layer be configured to first nerves network, in first nerves network to
Family sends sentence and is parsed, and obtains for representing that user sends the first intermediate vector of statement semantics；
S320：The decoding layer for talking with generation model is configured to nervus opticus network, to institute in the nervus opticus network
State intermediate vector to be parsed, obtain the vector group for representing sentence answer；And
S400：The vector group for representing sentence answer is exported as problem answers.
Wherein, when being parsed to the sentence that user sends in first nerves network in step S310, including following step
Suddenly：
S311：The sentence of user input is split into semantic minimum word unit in coding layer, obtains multiple words
Unit, and obtain the attribute of each word unit respectively, selects comprising at least one word more than information content as centre word, and by its
The input layer of first nerves network is input to as problem vector group in vector form；
S312：First nerves network hidden layer to described in the output of the input layer of first nerves network and previous moment
The output of the hidden layer of first nerves network carries out semantic parsing, and carries out linear weighted combination, forms the centre of representative sentences justice
Vector.
In step s 320, when being parsed to the intermediate vector in nervus opticus network, comprise the following steps：
S321：Received in decoding layer:Intermediate vector, and be input into intermediate vector as the input layer of nervus opticus network；
S322：It is refreshing to the intermediate vector from input layer and previous moment second in the hidden layer of nervus opticus network
Semantic parsing is carried out through the output of the hidden layer of network, some single vectors are sequentially generated, to form answer vector group, wherein answering
The semanteme of each single vector in case vector group corresponds to the semanteme of minimum word unit in answer output statement；
The answer vector group is exported in the output layer of nervus opticus network.
In another embodiment, after the answer vector group is exported as answer output statement, by the answer
Output statement is accordingly saved in knowledge base with dialogue read statement, is updated and is expanded with to knowledge base.
In another embodiment, after knowledge base matching primitives are carried out, whether there is and the dialogue according in knowledge base
The conversation sentence that the matching degree of read statement reaches predetermined value asks standard signal position to set, and according to request flag signal position
Validity come decide whether request dialogue generation model provide answer.
In another embodiment, linear weighted combination in step S312, comprises the following steps：
Step one：N groups center word data group is extracted in counting user read statement, wherein n is positive integer, in every group
The probability x that heart term data ZuλTian Nei centers word occursi, previous sentence center word data group λ Tian Nei centers word appearance
Probability yi, single argument regression model is set up,
yi=ω 'i·xi
Wherein, i is integer, i=1,2,3...... λ, ω 'iIt is the weighted regression coefficient in λ days；
Step 2：Solved using the least square method formula a kind of to step, the recurrence in λ days is calculated respectively
Coefficient estimate：
Wherein,It is the estimate of regression coefficient；xijFor jZu centers word data group ZhongiTian centers word occurs
Probability；It is the average value of j groups center word data group probability；yijIn in jth group user input sentence previous moment sentence
I-th day in heart term data group；The probability that center word occurs；It is putting down for j group previous moments center word data group probability
Average
Step 3：Normalized, the weighted value after being weighted：
Wherein, ωiIt is the weighted value after the weighting of user input sentence.
Preferably, the output collocation person of needing to use is selected and is used, and when exporting answer and being accurate, storage is being known
Know storehouse.
In another embodiment, the threshold speech noise reduction algorithm in step S100, including：
A, mute frame and speech frame are divided into by end-point detection by speech frame；
B, for mute frame, calculates the power spectral value of present frame as noise power spectrum estimate, for speech frame, calculates
Voice noise power Spectral Estimation value；
C, noise power spectrum estimate is subtracted by the power spectrum of speech frame, obtains the spectrum of the phonetic speech power after noise reduction；
D, the speech frame after noise reduction is drawn according to the phonetic speech power spectrum after noise reduction.
Voice noise power Spectral Estimation value computing formula is：
Wherein, I is noise power spectrum energy；Threshold valueN believes for noise
Number frame number；J=1-5It is conversion coefficient, e is natural constant；π is pi；fcIt is the frequency of noise signal；τ (t)=
0.03t2+0.6t+0.1；T is decomposition scale, 1≤t≤4.
I.e. by voice collection device, the noise collection of illustrative plates on voice is obtained, divided into speech frame by end-point detection
Mute frame and speech frame；For mute frame, the power spectral value of present frame is calculated as noise power spectrum estimate, for voice
Frame, calculates：
Wherein, I is noise power spectrum energy；Threshold valueN believes for noise
Number frame number；J=1-5It is conversion coefficient, e is natural constant；π is pi；fcIt is the frequency of noise signal；τ (t)=
0.03t2+0.6t+0.1；T is decomposition scale, 1≤t≤4.
Voice noise power Spectral Estimation value；The power spectrum of speech frame is subtracted into noise power spectrum estimate, after obtaining noise reduction
Phonetic speech power spectrum；Speech frame after noise reduction is drawn according to the phonetic speech power spectrum after noise reduction.
The present invention has designed and developed a kind of Machine self-learning based on neutral net and has built knowledge mapping training method, uses
Threshold speech noise reduction algorithm is obtained in that less mean square error, improves the signal to noise ratio of reconstructed speech signal, and using nerve
Network training generates dialog model, the model obtained using training, and what robot can be freely talks with user.
Although embodiment of the present invention is disclosed as above, it is not restricted to listed in specification and implementation method
With, it can be applied to various suitable the field of the invention completely, for those skilled in the art, can be easily
Other modification is realized, therefore under the universal limited without departing substantially from claim and equivalency range, the present invention is not limited
In specific details and shown here as the legend with description.
Claims (9)
1. a kind of Machine self-learning based on neutral net builds knowledge mapping training method, it is characterised in that including：
The sentence based on Nature condition that user sends is obtained, noise reduction is filtered to read statement using voice de-noising algorithm,
And obtain the classification of the sentence, and obtain the sentence above of the sentence, and the sentence above classification；
The sentence classification according to sentence, it is determined that the feedback sentence of matching；
If it does not exist, then according to neutral net dialog model, providing the answer of the sentence sent for user；Including：
The coding layer that the user sends statement model is configured to first nerves network, to user in the first nerves network
Send sentence to be parsed, obtain for representing that user sends the first intermediate vector of statement semantics；
The decoding layer of the dialogue generation model is configured to nervus opticus network, to the centre in the nervus opticus network
Vector is parsed, and obtains the vector group for representing sentence answer；And
The vector group for representing sentence answer is exported as problem answers.
2. the Machine self-learning based on neutral net according to claim 1 builds knowledge mapping training method, its feature
It is when being parsed to the sentence that the user sends in the first nerves network, to comprise the following steps：
The sentence of user input is split into semantic minimum word unit in coding layer, obtains multiple word units, and respectively
The attribute of each word unit is obtained, is selected comprising at least one word more than information content as centre word, and by it with vector
Form is input to the input layer of the first nerves network as problem vector group；
The first nerves network hidden layer to described in the output of the input layer of the first nerves network and previous moment
The output of the hidden layer of first nerves network carries out semantic parsing, and carries out linear weighted combination, forms the centre of representative sentences justice
Vector.
3. the Machine self-learning based on neutral net according to claim 2 builds knowledge mapping training method, its feature
It is when being parsed to the intermediate vector in the nervus opticus network, to comprise the following steps：
The intermediate vector is received in decoding layer, and is input into the intermediate vector as the input layer of nervus opticus network；
It is refreshing to described in the intermediate vector and previous moment from input layer second in the hidden layer of the nervus opticus network
Semantic parsing is carried out through the output of the hidden layer of network, some single vectors are sequentially generated, to form answer vector group, wherein institute
The semanteme for stating each single vector in answer vector group corresponds to the semanteme of minimum word unit in answer output statement；
The answer vector group is exported in the output layer of the nervus opticus network.
4. the Machine self-learning based on neutral net according to any one of claim 1-3 builds knowledge mapping training side
Method, it is characterised in that after the answer vector group is exported as answer output statement, by the answer output statement
Accordingly it is saved in knowledge base with dialogue read statement, is updated and expands with to knowledge base.
5. the Machine self-learning based on neutral net according to any one of claim 1-3 builds knowledge mapping training side
Method, it is characterised in that after knowledge base matching primitives are carried out, according to whether there is in knowledge base and the dialogue read statement
The conversation sentence that matching degree reaches predetermined value asks standard signal position to set, and next according to the validity of request flag signal position
Decide whether that request dialogue generation model provides answer.
6. the Machine self-learning based on neutral net according to claim 2 builds knowledge mapping training method, its feature
It is that the linear weighted combination is comprised the following steps：
Step one：N groups center word data group is extracted in counting user read statement, wherein n is positive integer, every group of centre word
The probability x that language data group λ Tian Nei centers word occursi, it is general that previous sentence center word data group λ Tian Nei centers word occurs
Rate yi, single argument regression model is set up,
yi=ω 'i·xi
Wherein, i is integer, i=1,2,3...... λ, ω 'iIt is the weighted regression coefficient in λ days；
Step 2：Solved using the least square method formula a kind of to step, the regression coefficient in λ days is calculated respectively
Estimate：
Wherein,It is the estimate of regression coefficient；xijFor jZu centers word data group ZhongiTian centers word occur it is general
Rate；It is the average value of j groups center word data group probability；yijCentre word in jth group user input sentence previous moment sentence
I-th day in language data group；The probability that center word occurs；It is the average value of j group previous moments center word data group probability
Step 3：Normalized, the weighted value after being weighted：
Wherein, ωiIt is the weighted value after the weighting of user input sentence.
7. the Machine self-learning based on neutral net according to claim 6 builds knowledge mapping training method, its feature
It is that output the collocation person of needing to use selected and used, when exporting answer and being accurate, storage is in knowledge base.
8. the Machine self-learning based on neutral net according to claim 1 builds knowledge mapping training method, its feature
It is, the voice de-noising algorithm, including：
A, mute frame and speech frame are divided into by end-point detection by speech frame；
B, for mute frame, calculates the power spectral value of present frame as noise power spectrum estimate, for speech frame, calculates voice
Noise power spectrum estimate；
C, noise power spectrum estimate is subtracted by the power spectrum of speech frame, obtains the spectrum of the phonetic speech power after noise reduction；
D, the speech frame after noise reduction is drawn according to the phonetic speech power spectrum after noise reduction.
9. the Machine self-learning based on neutral net according to claim 8 builds knowledge mapping training method, its feature
It is that the voice noise power Spectral Estimation value computing formula is：
Wherein, I is noise power spectrum energy；Threshold valueN is the frame of noise signal
Number；J=1-5It is conversion coefficient, e is natural constant；π is pi；fcIt is the frequency of noise signal；τ (t)=0.03t2+0.6t
+0.1；T is decomposition scale, 1≤t≤4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710127387.0A CN106875940B (en) | 2017-03-06 | 2017-03-06 | Machine self-learning construction knowledge graph training method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710127387.0A CN106875940B (en) | 2017-03-06 | 2017-03-06 | Machine self-learning construction knowledge graph training method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106875940A true CN106875940A (en) | 2017-06-20 |
| CN106875940B CN106875940B (en) | 2020-08-14 |
Family
ID=59171199
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710127387.0A Active CN106875940B (en) | 2017-03-06 | 2017-03-06 | Machine self-learning construction knowledge graph training method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106875940B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108108449A (en) * | 2017-12-27 | 2018-06-01 | 哈尔滨福满科技有限责任公司 | A kind of implementation method based on multi-source heterogeneous data question answering system and the system towards medical field |
| CN108389614A (en) * | 2018-03-02 | 2018-08-10 | 西安交通大学 | The method for building medical image collection of illustrative plates based on image segmentation and convolutional neural networks |
| WO2019024704A1 (en) * | 2017-08-03 | 2019-02-07 | 阿里巴巴集团控股有限公司 | Entity annotation method, intention recognition method and corresponding devices, and computer storage medium |
| CN109933773A (en) * | 2017-12-15 | 2019-06-25 | 上海擎语信息科技有限公司 | A kind of multiple semantic sentence analysis system and method |
| CN110349463A (en) * | 2019-07-10 | 2019-10-18 | 南京硅基智能科技有限公司 | A kind of reverse tutoring system and method |
| CN112309183A (en) * | 2020-11-12 | 2021-02-02 | 江苏经贸职业技术学院 | Interactive listening and speaking exercise system suitable for foreign language teaching |
| CN112487173A (en) * | 2020-12-18 | 2021-03-12 | 北京百度网讯科技有限公司 | Man-machine conversation method, device and storage medium |
| CN112528039A (en) * | 2020-12-16 | 2021-03-19 | 中国联合网络通信集团有限公司 | Word processing method, device, equipment and storage medium |
| CN113316752A (en) * | 2019-01-24 | 2021-08-27 | 索尼半导体解决方案公司 | Voltage control device |
| WO2021190389A1 (en) * | 2020-03-25 | 2021-09-30 | 阿里巴巴集团控股有限公司 | Speech processing method, speech encoder, speech decoder, and speech recognition system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1647069A (en) * | 2002-04-11 | 2005-07-27 | 株式会社PtoPA | Conversation control system and conversation control method |
| US20060020473A1 (en) * | 2004-07-26 | 2006-01-26 | Atsuo Hiroe | Method, apparatus, and program for dialogue, and storage medium including a program stored therein |
| CN104217226A (en) * | 2014-09-09 | 2014-12-17 | 天津大学 | Dialogue act identification method based on deep neural networks and conditional random fields |
| CN105704013A (en) * | 2016-03-18 | 2016-06-22 | 北京光年无限科技有限公司 | Context-based topic updating data processing method and apparatus |
| CN105787560A (en) * | 2016-03-18 | 2016-07-20 | 北京光年无限科技有限公司 | Dialogue data interaction processing method and device based on recurrent neural network |
| CN106055662A (en) * | 2016-06-02 | 2016-10-26 | 竹间智能科技（上海）有限公司 | Emotion-based intelligent conversation method and system |
-
2017
- 2017-03-06 CN CN201710127387.0A patent/CN106875940B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1647069A (en) * | 2002-04-11 | 2005-07-27 | 株式会社PtoPA | Conversation control system and conversation control method |
| US20060020473A1 (en) * | 2004-07-26 | 2006-01-26 | Atsuo Hiroe | Method, apparatus, and program for dialogue, and storage medium including a program stored therein |
| CN104217226A (en) * | 2014-09-09 | 2014-12-17 | 天津大学 | Dialogue act identification method based on deep neural networks and conditional random fields |
| CN105704013A (en) * | 2016-03-18 | 2016-06-22 | 北京光年无限科技有限公司 | Context-based topic updating data processing method and apparatus |
| CN105787560A (en) * | 2016-03-18 | 2016-07-20 | 北京光年无限科技有限公司 | Dialogue data interaction processing method and device based on recurrent neural network |
| CN106055662A (en) * | 2016-06-02 | 2016-10-26 | 竹间智能科技（上海）有限公司 | Emotion-based intelligent conversation method and system |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019024704A1 (en) * | 2017-08-03 | 2019-02-07 | 阿里巴巴集团控股有限公司 | Entity annotation method, intention recognition method and corresponding devices, and computer storage medium |
| CN109388793A (en) * | 2017-08-03 | 2019-02-26 | 阿里巴巴集团控股有限公司 | Entity mask method, intension recognizing method and corresponding intrument, computer storage medium |
| CN109388793B (en) * | 2017-08-03 | 2023-04-07 | 阿里巴巴集团控股有限公司 | Entity marking method, intention identification method, corresponding device and computer storage medium |
| CN109933773B (en) * | 2017-12-15 | 2023-05-26 | 上海擎语信息科技有限公司 | Multiple semantic statement analysis system and method |
| CN109933773A (en) * | 2017-12-15 | 2019-06-25 | 上海擎语信息科技有限公司 | A kind of multiple semantic sentence analysis system and method |
| CN108108449A (en) * | 2017-12-27 | 2018-06-01 | 哈尔滨福满科技有限责任公司 | A kind of implementation method based on multi-source heterogeneous data question answering system and the system towards medical field |
| CN108389614B (en) * | 2018-03-02 | 2021-01-19 | 西安交通大学 | Method for constructing medical image map based on image segmentation and convolutional neural network |
| CN108389614A (en) * | 2018-03-02 | 2018-08-10 | 西安交通大学 | The method for building medical image collection of illustrative plates based on image segmentation and convolutional neural networks |
| CN113316752A (en) * | 2019-01-24 | 2021-08-27 | 索尼半导体解决方案公司 | Voltage control device |
| CN110349463A (en) * | 2019-07-10 | 2019-10-18 | 南京硅基智能科技有限公司 | A kind of reverse tutoring system and method |
| WO2021190389A1 (en) * | 2020-03-25 | 2021-09-30 | 阿里巴巴集团控股有限公司 | Speech processing method, speech encoder, speech decoder, and speech recognition system |
| CN112309183A (en) * | 2020-11-12 | 2021-02-02 | 江苏经贸职业技术学院 | Interactive listening and speaking exercise system suitable for foreign language teaching |
| CN112528039A (en) * | 2020-12-16 | 2021-03-19 | 中国联合网络通信集团有限公司 | Word processing method, device, equipment and storage medium |
| CN112487173A (en) * | 2020-12-18 | 2021-03-12 | 北京百度网讯科技有限公司 | Man-machine conversation method, device and storage medium |
| CN112487173B (en) * | 2020-12-18 | 2021-09-10 | 北京百度网讯科技有限公司 | Man-machine conversation method, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106875940B (en) | 2020-08-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106875940A (en) | A kind of Machine self-learning based on neutral net builds knowledge mapping training method | |
| CN108536681B (en) | Intelligent question-answering method, device, equipment and storage medium based on emotion analysis | |
| CN108734276B (en) | Simulated learning dialogue generation method based on confrontation generation network | |
| CN105787560B (en) | Dialogue data interaction processing method and device based on Recognition with Recurrent Neural Network | |
| CN109036465B (en) | Speech emotion recognition method | |
| CN106095950B (en) | Professor is intended to answer generation method in a kind of human-computer dialogue | |
| Rothe et al. | Question asking as program generation | |
| CN108829756B (en) | Method for solving multi-turn video question and answer by using hierarchical attention context network | |
| CN106779053A (en) | The knowledge point of a kind of allowed for influencing factors and neutral net is known the real situation method | |
| CN109710744A (en) | A kind of data matching method, device, equipment and storage medium | |
| CN110457661A (en) | Spatial term method, apparatus, equipment and storage medium | |
| CN113761156A (en) | Data processing method, device and medium for man-machine interaction conversation and electronic equipment | |
| CN115393933A (en) | Video face emotion recognition method based on frame attention mechanism | |
| Platonov et al. | A spoken dialogue system for spatial question answering in a physical blocks world | |
| Liu et al. | Perspective-corrected spatial referring expression generation for human–robot interaction | |
| Adewale et al. | Pixie: a social chatbot | |
| CN109190116A (en) | Semantic analytic method, system, electronic equipment and storage medium | |
| Yang et al. | Multi-intent text classification using dual channel convolutional neural network | |
| CN114168769B (en) | Visual question-answering method based on GAT relation reasoning | |
| CN113761149A (en) | Dialogue information processing method, device, computer equipment and storage medium | |
| Katsumi et al. | Optimization of information-seeking dialogue strategy for argumentation-based dialogue system | |
| CN114625986A (en) | Method, device and equipment for sorting search results and storage medium | |
| CN115408500A (en) | Question-answer consistency evaluation method and device, electronic equipment and medium | |
| Singh et al. | Human perception based criminal identification through human robot interaction | |
| CN115186072A (en) | Knowledge graph visual question-answering method based on double-process cognitive theory |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Wang Dongliang Inventor after: Liu Yingbo Inventor after: Wang Hongbin Inventor after: Jiang Yuji Inventor after: Yao Xing Inventor after: Yu Yanlong Inventor after: Li Xiaowen Inventor after: Zhang Zhiwei Inventor after: Zhang Lingling Inventor before: Liu Yingbo Inventor before: Wang Dongliang Inventor before: Wang Hongbin |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |