CN108364066B - Artificial neural network chip and its application method based on N-GRAM and WFST model - Google Patents
Artificial neural network chip and its application method based on N-GRAM and WFST model Download PDFInfo
- Publication number
- CN108364066B CN108364066B CN201810249286.5A CN201810249286A CN108364066B CN 108364066 B CN108364066 B CN 108364066B CN 201810249286 A CN201810249286 A CN 201810249286A CN 108364066 B CN108364066 B CN 108364066B
- Authority
- CN
- China
- Prior art keywords
- neural network
- attention
- score
- model
- reply message
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9032—Query formulation
- G06F16/90332—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Human Computer Interaction (AREA)
- Neurology (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Present disclose provides a kind of automatic chatting methods based on deep neural network, comprising the following steps: obtains user and inputs information, and generates regular length vector through deep neural network encoder；Score is exported after the regular length vector input attention model；Determined to generate reply message corresponding with the input information via attention model or natural model according to the score；Wherein, if the score enters language model less than a score threshold, reply message corresponding with the input information is generated through language model；Conversely, directly generating reply message corresponding with the input information via attention model.The disclosure additionally provides a kind of automatic chatting robot based on deep neural network.Automatic chatting method and robot of the disclosure based on deep neural network ensure that the accuracy of reply message in chat process, so that chat content is more realistic.
Description
Technical field
This disclosure relates to field of artificial intelligence more particularly to a kind of automatic chatting method based on deep neural network
And robot.
Background technique
In recent years, with the rapid development of artificial intelligence, chat robots also receive the extensive of academia and industry
Concern.
But traditional chat robots are there is also being faced with many problems, such as dialogue can not generate and specifically contain
Justice, context do not accept logic, are unable to satisfy use demand.
In particular, China is rapidly entering aging society at present.In future society, how to support parents will be one it is huge
Big society and economy problem.The mode that present old-age provision model mostly uses nursing staff that old man is accompanied to chat, to be reached for
Old man, which provides, accompanies chatting service, provides the affectional consolation of the elderly.And it is artificial go to chat with the elderly will occupy it is big
The manpower and material resources of amount, if this service that will chat with the elderly is completed using chat robots, it will have it is huge economical and
Social benefit.However traditional chat robots can not meet the need for affection of the elderly well.The elderly can have found oneself
Talk with machine, to not like these chats.
Summary of the invention
(1) technical problems to be solved
In order to solve or at least partly alleviate above-mentioned technical problem, present disclose provides one kind to be based on deep neural network
Automatic chatting method and robot.
(2) technical solution
According to one aspect of the disclosure, a kind of automatic chatting method based on deep neural network is provided, including with
Lower step: it obtains user and inputs information, and generate regular length vector through deep neural network encoder；The regular length to
Score is exported after amount input attention model；And determined according to the score via attention model or natural norm
Type generates reply message corresponding with the input information；Wherein, if the score enters language mould less than a score threshold
Type generates reply message corresponding with the input information through language model；Conversely, directly being generated via attention model
Reply message corresponding with the input information.
In some embodiments, the attention model is the depth nerve net for including one or more neural net layers
Network, export after regular length vector input attention model all may score corresponding to reply message, select
Reply message corresponding to highest score is as final reply message.
It in some embodiments, include full articulamentum and softmax layers in the attention model；It is described fixed long
Degree vector makees vector/matrix multiplication and/or add operation in attention model to export the score.
In some embodiments, the attention model includes multiple attention submodels, and difference is respectively adopted
The training of classification corpus forms；Multiple attention submodel is connected in parallel, and regular length vector inputs the multiple respectively
Multiple scores are exported after neural network in attention submodel, if the highest score in the multiple score is less than one point
Number threshold value, then enter language model, generates reply message corresponding with the input information through language model；Conversely, directly passing through
Attention model generates reply message corresponding with the input information.
In some embodiments, the multiple attention submodel includes: the first attention submodel comprising
Full articulamentum neural network, softmax layers of neural network and convolutional layer neural network, are instructed using daily life classification corpus
Practice；2nd attention submodel comprising norm layers of convolutional layer, full articulamentum, pond layer and batch neural network are adopted
It is trained with current events news category corpus；3rd attention submodel comprising full articulamentum neural network, pond layer
Neural network, softmax layer neural network and convolutional layer neural network are trained using soul emotional category corpus.
In some embodiments, the attention model includes multiple attention submodels；Via
When attention model generates reply message corresponding with the input information, the multiple attention submodel is exported
Score be compared, select attention submodel corresponding to highest score as generate reply message final mask.
In some embodiments, the language model is generated using N-GRAM, WFST；In language model, searched using A*
Rope, beam search generate reply message corresponding with the input information.
In some embodiments, the natural model generates reply message in the form of text, and language model is each time
The new score of all texts to be selected is generated in iteration, and all texts in a time series thus can be generated after the completion of iteration
Search space obtains final unique output result according to the regular length vector that encoder generates in the search space.
A kind of automatic chatting robot based on deep neural network another aspect of the present disclosure provides, packet
Include: preprocessing module, for obtain user input information, and by a deep neural network encoder to the input information into
Row pretreatment, generates regular length vector；Processing module: for receiving the regular length vector, and attention mould is utilized
Type exports score；And determine and generation module, for according to the score determine via attention model or natural model
Generate reply message corresponding with the input information；Wherein, if the score enters language mould less than a score threshold
Type generates reply message corresponding with the input information through language model；Conversely, directly being generated via attention model
Reply message corresponding with the input information.
In some embodiments, the attention model is the depth nerve net for including one or more neural net layers
Network, export after regular length vector input attention model all may score corresponding to reply message, select
Reply message corresponding to highest score is as final reply message.
(3) beneficial effect
It can be seen from the above technical proposal that automatic chatting method and robot of the disclosure based on deep neural network are extremely
One of them is had the advantages that less:
(1) due to the artificial automatic chatting robot based on deep neural network of disclosure automatic chatting method and machine,
Thus it can ensure that the accuracy of reply message in chat process by updating weight in training process, keep dialogue trueer
It is real.
(2) disclosure is due to disclosure automatic chatting method and robot, according to the score of attention model output or
Determine to generate reply message corresponding with the input information via attention model or natural model, has fully considered and chatted
The particular content of information is inputted during it, and its common and uncommon property is distinguished, and different models is then respectively adopted
Reply message is generated, so that chat content is more realistic.
(3) with artificial intelligence technology, the especially development of deep neural network technology, intelligent chat robots are just gradually
More and more actively in people's daily life, traditional artificial company chat is passed through newest artificial intelligence by the disclosure
Counting automates it, and effective solution cost of labor while brings high economic benefit, needs people to come the tradition
At task by technological means impart computer chat function.
(4) disclosure has fully considered the chat content feature of chatting object group, such as this group of the elderly,
Information can be inputted to it in conjunction with the characteristics of the elderly's chat to distinguish, introduce attention model and language model is total
With the regular length vector after processing LSTM Processing with Neural Network；Using attention model treatment the elderly chat in relatively often
The corpus of appearance using the corpus of less appearance in language model processing the elderly's chat, and uses score as using
The judgement of attention model or language model, so that chat content is more realistic；It can accomplish to contain in chat process
The function of the concrete meanings such as logic more meets the need for affection of the elderly, them is made to be ready to talk with chat robots.
Detailed description of the invention
Fig. 1 is disclosure neural network training process schematic diagram.
Fig. 2 is disclosure automatic chatting method flow chart.
Fig. 3 is disclosure automatic chatting robot block diagram.
Specific embodiment
For the purposes, technical schemes and advantages of the disclosure are more clearly understood, below in conjunction with specific embodiment, and reference
Attached drawing is described in further detail the disclosure.
It should be noted that similar or identical part all uses identical figure number in attached drawing or specification description.It is attached
The implementation for not being painted or describing in figure is form known to a person of ordinary skill in the art in technical field.In addition, though this
Text can provide the demonstration of the parameter comprising particular value, it is to be understood that parameter is equal to corresponding value without definite, but can connect
It is similar to be worth accordingly in the error margin or design constraint received.In addition, the direction term mentioned in following embodiment, such as
"upper", "lower", "front", "rear", "left", "right" etc. are only the directions with reference to attached drawing.Therefore, the direction term used be for
Illustrate not to be used to limit the disclosure.
It is described based on deep neural network present disclose provides a kind of automatic chatting method based on deep neural network
Automatic chatting method the following steps are included:
S1 obtains user and inputs information, and generates regular length vector through deep neural network encoder；
S2, the regular length vector input attention model export score later；And
S3 determines to generate via attention model or natural model corresponding with the input information according to the score
Reply message.
Wherein, it if the score enters language model less than a score threshold, is generated and the input through language model
The corresponding reply message of information；Believe conversely, directly generating answer corresponding with the input information via attention model
Breath.
It can be the deep neural network for including one or more neural net layers in the attention model.It is described solid
Measured length vector input after attention model export all may score corresponding to reply message, selection highest score
Corresponding reply message is as final reply message.It optionally, include full articulamentum nerve net in the attention model
Network and softmax layers of neural network；The regular length vector is made vector/matrix multiplication and/or is added in attention model
Method operation is to export the score.
Further, the attention model includes multiple attention submodels, and different classes of language is respectively adopted
Material training forms；Multiple attention submodel is connected in parallel, and regular length vector inputs the multiple respectively
Multiple scores are exported after neural network in attention submodel, if the highest score in the multiple score is less than one point
Number threshold value, then enter language model, generates reply message corresponding with the input information through language model；Conversely, directly passing through
Attention model generates reply message corresponding with the input information.
Each submodel that the attention model includes is trained using different classes of corpus set, these are not
Generic language chats set can be according to being determined using the chat content statistical result of object to chat robots.Such as to more
It is a to be counted using the chat content of object within a certain period of time, determine most common multiple corpus classifications.Citing comes
It says, the multiple attention submodel can include: the first attention submodel comprising full articulamentum neural network,
Softmax layers of neural network and convolutional layer neural network are trained using daily life classification corpus；2nd attention
Submodel comprising norm layers of convolutional layer, full articulamentum, pond layer and batch neural network, using topical news classification language
Material is trained；3rd attention submodel comprising full articulamentum neural network, pond layer neural network, softmax
Layer neural network and convolutional layer neural network, are trained using soul emotional category corpus.If the attention model packet
Multiple attention submodels are included, then generating reply message corresponding with the input information via attention model
When, dividing for can exporting to the multiple attention submodel is compared, and is selected corresponding to highest score
Attention submodel is as the final mask for generating reply message.As a result, by being set to each attention model structure
It sets, the final mask of the setting and reply message of training corpus set is selected, so that the reply message generated is more acurrate, more
Meet reality, better meets the demand using object.
Specifically, generating the language model using N-GRAM, WFST；In language model, A* search, beam search are used
Generate reply message corresponding with the input information.
More specifically, the natural model generates reply message in the form of text, and language model is in iteration each time
It is middle to generate the new score of all texts to be selected, the search of all texts in a time series thus can be generated after the completion of iteration
Space obtains final unique output result according to the regular length vector that encoder generates in the search space.
Below by taking this applicable object group of the elderly as an example, the disclosure is discussed in detail based on the automatic of deep neural network
Chat process.
Deep neural network algorithm is divided into training process and use process two parts.In the training process, usage history is received
The text information data of the elderly's chat data of collection, i.e. the elderly and true people dialogue are as training set training book depth nerve
Network is helped the elderly chat robots.Specifically: input: the natural language of the written form of the elderly user's input.Intermediate treatment: it compiles
Code device coding and decoder decoding；Output and update: it is compared according to output result and legitimate reading, updates the power of neural network
Value.
Wherein, encoder receives the vector of regular length that user inputs and generates, and the vector of regular length enters decoding
Device decoding generates purpose and replys in language dictionary, the probability of each word, it is however generally that, there are several words in corpus just and have and is several
A neuron is indicated score by number in each neuron, and the total number of neuron is certain, therefore the output of network is just
For the score of each neuron.The encoder can be including convolutional layer, full articulamentum, pond layer and norm layers of batch
LSTM neural network.This LSTM (shot and long term memory network) neural network combination chat feature, i.e., when in the chat of last sentence
Appearance is this feature relevant with a few words before, devises corresponding LSTM neural network, i.e. present networks will can go out recently
In feeding LSTM neural network of the existing input as secondary input iteration.As shown in Figure 1, in the training process of neural network
Input language is divided into X in time sequencingt-1, Xt, Xt+1Three segments, and it is i.e. solid to generate corresponding neural network result to every a word
Vector (the h of measured lengtht-1、ht、ht+1) be sent to the words as output in next iteration simultaneously.That is previous in Fig. 1
Box can all be introduced into the input in next box, and (three box internal structures can be identical in Fig. 1, i.e., unshowned first party
Frame and third box internal structure can be identical as the internal structure shown in the second box).In the training process, the nerve net obtained
The error that the output of network and pre-prepd data set ask absolute average error (AAE) or minimum mean-square error (LSE) to generate
Gradient is generated by direction propagation algorithm (BP).The weight in gradient updating LSTM is used in training process.Continuous iteration should
The part chat robots LSTM of actual use is generated after LSTM neural network reduction error.
The more situation of term is repeated in addition, chatting in conjunction with the elderly, such as the deeper thing of name or impression of sons and daughters
Part, these language can occur repeatedly in one section of event in the dialogue of the elderly.Therefore this chat robots introduces
Attention model carries out the chat sentence relatively often having separately encoded.Attention model realization it is specific it is certain compared with
Often have chat sentence corresponding relationship, specifically, the realization by full articulamentum neural network, softmax layers of neural network,
Matrix multiplication, addition of matrices are completed.
Attention model be for the daily chat of the elderly common language be target deep neural network.The mind
Through in network structure a softmax network layer can be followed by for full articulamentum neural network.On specifically used, in advance to the elderly
Daily chat situation is counted, and the chat corresponding relationship often occurred in the elderly's chat process is obtained.Use the data pair
Attention model is trained.Different corpus set trains multiple attention models.Such as using talking about parent
In phrase material, talk about daily life corpus, talk about topical news corpus and can train three different attention models
Above-mentioned sequence of maneuvers is defined as attention, can mutually be contacted with multiple attention, in parallel, great-jump-forward string
The permutation and combination methods such as company are new attention.Specifically: the elderly's different classes of language in term of speaking is trained
Different attention models.Such as: a full articulamentum, which is trained, using the corpus for talking about neighborhood relationship is followed by softmax
The neural network attention1 of layer, trains a neural network using the corpus for talking about breakfast lunch supper
Attention2 trains a neural network attention3 using the corpus for talking about relatives and friends.Specifically used upper one
A old man's word language can enter simultaneously after the regular length vector that above-mentioned neural network encoder generates
Attention1, attention2, attention3 generate three corresponding responses, are selected according to the score of each response last
Attention model corresponds to response.
For content more uncommon in the elderly's session request, possibly correspondence can not be found from attention model
Chat response coding after vector, i.e., the score acquired in above-mentioned attention model is less than some threshold value or acquires
Gauss distance be greater than some threshold value when, that is, determine the content be uncommon content.Language model is introduced for these contents to deposit
It stores up and generates target retro result in conjunction with the sort algorithms such as priori knowledge and beam search, A* algorithm.In preparatory natural language processing
Priori knowledge establish corresponding language model.By the score of all texts to be selected of generation of language model iteration, each
The new score of all texts to be selected is generated in secondary iteration, and all texts in a time series thus can be generated after the completion of iteration
Search space, the vector of the regular length generated within this space according to encoder obtains final unique language output result.
Specifically, can realize language model by WFST n-gram algorithm, the natural language of reply is generated later, with the side such as text
Formula exports out.The language can find out error with corresponding language in training set in the training process.The error of generation via
Each neuron that BP (back-propagation algorithm) returns to neural network corresponds to the weight of neuron as gradient updating.The process
By iterating, weight constantly with new neural network is the knot in result and training set that last neural network generates
Fruit is close, then completes the training of neural network, and the neural network is exactly a chat robots of completely helping the elderly after training.
For the course of work of the chat robots, as shown in Figure 2.Specifically: the elderly input written form from
Right language.Intermediate treatment: intermediate treatment process is that the natural language of the written form of the elderly's input of input first passes through LSTM
After neural network encoder generates the vector of a regular length, is generated by attention model, language model and reply letter
Breath.
To the uncertain natural language phrase of word length, this phrase is usually the elderly's chat pair that written form indicates
A word that the elderly in words says.The phrase generates after previously described trained LSTM neural network in advance
The fixed vector of one length.The vector enters in attention model later.If parallel into several simultaneously
After attention model, several are generated by the neural network in attention model and corresponds to output result and score,
Select the highest one group of result of score as output.
If all scores in above-mentioned multiple scores are below some threshold value.The then fixation that LSTM neural network generates
The vector (result generated without using attention model) of length, into language model, by above utilizing N-
The corresponding chat of the phrase is generated using searching algorithms such as A* search, beam searches in the language model that GRAM, WFST technology generate
Response phrase.It generates to come finally by the form of text, the response of this dialogue as the elderly.The elderly next time
Dialogue will be again inputted into LSTM neural network model, carry out an above process.Thus iteration uses this chatting machine
Device people, the generation dialogue of question-response, the elderly (user) is asking that chat robots are being answered, final to realize that the elderly helps the elderly merely
It function.
The disclosure additionally provides a kind of automatic chatting robot based on deep neural network, as shown in figure 3, described automatic
Chat robots, comprising:
Preprocessing module inputs information for obtaining user, and by a deep neural network encoder to the input
Information is pre-processed, and regular length vector is generated；
Processing module: score is exported for receiving the regular length vector, and using attention model；And
Determine and generation module, for according to the score determine to generate via attention model or natural model with
The corresponding reply message of the input information.
Wherein, it if the score enters language model less than a score threshold, is generated and the input through language model
The corresponding reply message of information；Believe conversely, directly generating answer corresponding with the input information via attention model
Breath.
The attention model is the deep neural network for including one or more neural net layers, the regular length
Vector input after attention model export all may score corresponding to reply message, corresponding to selection highest score
Reply message as final reply message.
Particular embodiments described above has carried out further in detail the purpose of the disclosure, technical scheme and beneficial effects
Describe in detail it is bright, it is all it should be understood that be not limited to the disclosure the foregoing is merely the specific embodiment of the disclosure
Within the spirit and principle of the disclosure, any modification, equivalent substitution, improvement and etc. done should be included in the guarantor of the disclosure
Within the scope of shield.
Claims (9)
1. a kind of automatic chatting method based on deep neural network, comprising the following steps:
It obtains user and inputs information, and generate regular length vector through deep neural network encoder；
Score is exported after the regular length vector input attention model；And
Determined to generate reply message corresponding with the input information via attention model according to the score；Wherein, if
The score then enters language model less than a score threshold, generates answer corresponding with the input information through language model
Information；Conversely, directly generating reply message corresponding with the input information via attention model；
Wherein, the attention model includes multiple attention submodels, the multiple attention submodel packet
It includes:
First attention submodel comprising full articulamentum neural network, softmax layers of neural network and convolutional layer nerve
Network is trained using daily life classification corpus；
2nd attention submodel comprising norm layers of convolutional layer, full articulamentum, pond layer and batch neural network are adopted
It is trained with current events news category corpus；
3rd attention submodel comprising full articulamentum neural network, pond layer neural network, softmax layers of nerve net
Network and convolutional layer neural network are trained using soul emotional category corpus.
2. the automatic chatting method according to claim 1 based on deep neural network, wherein the attention mould
Type is the deep neural network for including one or more neural net layers, regular length vector input attention model it
Score corresponding to whole reply messages is exported afterwards, selects reply message corresponding to highest score as final reply message.
3. the automatic chatting method according to claim 2 based on deep neural network, wherein the attention mould
It include full articulamentum and softmax layers in type；
The regular length vector makees vector/matrix multiplying in attention model to export the score；Or
The regular length vector makees vector/matrix add operation in attention model to export the score；Or
The regular length vector make in attention model vector/matrix multiplying and vector/matrix add operation from
And export the score.
4. the automatic chatting method according to claim 1 based on deep neural network, wherein the attention mould
The multiple attention submodel that type includes is respectively adopted different classes of corpus training and forms；It is the multiple
Attention submodel is connected in parallel, and regular length vector inputs the nerve net in the multiple attention submodel respectively
Multiple scores are exported after network, if the highest score in the multiple score enters language model, through language less than a score threshold
Say that model generates reply message corresponding with the input information；Conversely, directly being generated and the input through attention model
The corresponding reply message of information.
5. the automatic chatting method according to claim 1 based on deep neural network, wherein the attention mould
The multiple attention submodel that type includes；Answer corresponding with the input information is being generated via attention model
When complex information, the score of the multiple attention submodel output is compared, is selected corresponding to highest score
Attention submodel is as the final mask for generating reply message.
6. the automatic chatting method according to claim 1 based on deep neural network, wherein utilize N-GRAM, WFST
Generate the language model；In language model, answer corresponding with the input information is generated using A* search, beam search and is believed
Breath.
7. the automatic chatting method according to claim 1 based on deep neural network, wherein the language model is every
The new score of all texts to be selected is generated in an iteration, and all texts in a time series thus can be generated after the completion of iteration
The search space of word obtains final unique output knot according to the regular length vector that encoder generates in the search space
Fruit.
8. a kind of automatic chatting robot based on deep neural network, comprising:
Preprocessing module inputs information for obtaining user, and by a deep neural network encoder to the input information
It is pre-processed, generates regular length vector；
Processing module: score is exported for receiving the regular length vector, and using attention model；And
Judgement and generation module, for determining to generate and the input information pair via attention model according to the score
The reply message answered；Wherein, if the score enters language model less than a score threshold, through language model generation and institute
State the corresponding reply message of input information；Conversely, directly generating answer corresponding with the input information via attention model
Complex information；
Wherein, the attention model includes multiple attention submodels, the multiple attention submodel packet
It includes:
First attention submodel comprising full articulamentum neural network, softmax layers of neural network and convolutional layer nerve
Network is trained using daily life classification corpus；
2nd attention submodel comprising norm layers of convolutional layer, full articulamentum, pond layer and batch neural network are adopted
It is trained with current events news category corpus；
3rd attention submodel comprising full articulamentum neural network, pond layer neural network, softmax layers of nerve net
Network and convolutional layer neural network are trained using soul emotional category corpus.
9. the automatic chatting robot according to claim 8 based on deep neural network, wherein the attention
Model is the deep neural network for including one or more neural net layers, and the regular length vector inputs attention model
Score corresponding to whole reply messages is exported later, and reply message corresponding to highest score is selected to reply letter as final
Breath.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2017112403736 | 2017-11-30 | ||
| CN201711240373 | 2017-11-30 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108364066A CN108364066A (en) | 2018-08-03 |
| CN108364066B true CN108364066B (en) | 2019-11-08 |
Family
ID=63001070
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810249286.5A Active CN108364066B (en) | 2017-11-30 | 2018-03-23 | Artificial neural network chip and its application method based on N-GRAM and WFST model |
| CN201810249289.9A Active CN108388944B (en) | 2017-11-30 | 2018-03-23 | A kind of automatic chatting method and robot based on deep neural network |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810249289.9A Active CN108388944B (en) | 2017-11-30 | 2018-03-23 | A kind of automatic chatting method and robot based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN108364066B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255020B (en) * | 2018-09-11 | 2022-04-01 | 浙江大学 | Method for solving dialogue generation task by using convolution dialogue generation model |
| CN109901878B (en) * | 2019-02-25 | 2021-07-23 | 北京灵汐科技有限公司 | Brain-like computing chip and computing equipment |
| CN110309287B (en) * | 2019-07-08 | 2021-07-06 | 北京邮电大学 | Retrieval type chatting dialogue scoring method for modeling dialogue turn information |
| CN110491394B (en) * | 2019-09-12 | 2022-06-17 | 北京百度网讯科技有限公司 | Awakening corpus obtaining method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503066A (en) * | 2016-09-29 | 2017-03-15 | 北京百度网讯科技有限公司 | Process Search Results method and apparatus based on artificial intelligence |
| CN106682387A (en) * | 2016-10-26 | 2017-05-17 | 百度国际科技（深圳）有限公司 | Method and device used for outputting information |
| CN107133209A (en) * | 2017-03-29 | 2017-09-05 | 北京百度网讯科技有限公司 | Comment generation method and device, equipment and computer-readable recording medium based on artificial intelligence |
| EP3229157A1 (en) * | 2016-04-07 | 2017-10-11 | Siemens Healthcare GmbH | Image analytics question answering |
| CN107329996A (en) * | 2017-06-08 | 2017-11-07 | 三峡大学 | A kind of chat robots system and chat method based on fuzzy neural network |
| CN107358948A (en) * | 2017-06-27 | 2017-11-17 | 上海交通大学 | Language in-put relevance detection method based on attention model |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104809103B (en) * | 2015-04-29 | 2018-03-30 | 北京京东尚科信息技术有限公司 | A kind of interactive semantic analysis and system |
| US20160350653A1 (en) * | 2015-06-01 | 2016-12-01 | Salesforce.Com, Inc. | Dynamic Memory Network |
| CN107220641B (en) * | 2016-03-22 | 2020-06-26 | 华南理工大学 | Multi-language text classification method based on deep learning |
| CN106776578B (en) * | 2017-01-03 | 2020-03-17 | 竹间智能科技（上海）有限公司 | Method and device for improving conversation performance of conversation system |
| CN106951468B (en) * | 2017-03-02 | 2018-12-28 | 腾讯科技（深圳）有限公司 | Talk with generation method and device |
| CN107025283A (en) * | 2017-04-05 | 2017-08-08 | 竹间智能科技（上海）有限公司 | The answer method and system of candidate answers sequence are carried out based on subscriber data |
| CN107273487A (en) * | 2017-06-13 | 2017-10-20 | 北京百度网讯科技有限公司 | Generation method, device and the computer equipment of chat data based on artificial intelligence |
| CN107562792B (en) * | 2017-07-31 | 2020-01-31 | 同济大学 | question-answer matching method based on deep learning |
| CN107766506A (en) * | 2017-10-20 | 2018-03-06 | 哈尔滨工业大学 | A kind of more wheel dialog model construction methods based on stratification notice mechanism |
| CN107798126B (en) * | 2017-11-13 | 2021-11-02 | 北京邮电大学 | Question-answer processing method based on knowledge base |
-
2018
- 2018-03-23 CN CN201810249286.5A patent/CN108364066B/en active Active
- 2018-03-23 CN CN201810249289.9A patent/CN108388944B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3229157A1 (en) * | 2016-04-07 | 2017-10-11 | Siemens Healthcare GmbH | Image analytics question answering |
| CN106503066A (en) * | 2016-09-29 | 2017-03-15 | 北京百度网讯科技有限公司 | Process Search Results method and apparatus based on artificial intelligence |
| CN106682387A (en) * | 2016-10-26 | 2017-05-17 | 百度国际科技（深圳）有限公司 | Method and device used for outputting information |
| CN107133209A (en) * | 2017-03-29 | 2017-09-05 | 北京百度网讯科技有限公司 | Comment generation method and device, equipment and computer-readable recording medium based on artificial intelligence |
| CN107329996A (en) * | 2017-06-08 | 2017-11-07 | 三峡大学 | A kind of chat robots system and chat method based on fuzzy neural network |
| CN107358948A (en) * | 2017-06-27 | 2017-11-17 | 上海交通大学 | Language in-put relevance detection method based on attention model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108388944B (en) | 2019-10-18 |
| CN108364066A (en) | 2018-08-03 |
| CN108388944A (en) | 2018-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108364066B (en) | Artificial neural network chip and its application method based on N-GRAM and WFST model | |
| CN108875807B (en) | Image description method based on multiple attention and multiple scales | |
| CN104598611B (en) | The method and system being ranked up to search entry | |
| CN107562863A (en) | Chat robots reply automatic generation method and system | |
| CN109241255A (en) | A kind of intension recognizing method based on deep learning | |
| CN108363690A (en) | Dialog semantics Intention Anticipation method based on neural network and learning training method | |
| CN107330444A (en) | A kind of image autotext mask method based on generation confrontation network | |
| CN109783817A (en) | A kind of text semantic similarity calculation model based on deeply study | |
| CN107301168A (en) | Intelligent robot and its mood exchange method, system | |
| CN110222163A (en) | A kind of intelligent answer method and system merging CNN and two-way LSTM | |
| CN106503055A (en) | A kind of generation method from structured text to iamge description | |
| CN107679225B (en) | Reply generation method based on keywords | |
| CN106682387A (en) | Method and device used for outputting information | |
| CN110059191A (en) | A kind of text sentiment classification method and device | |
| CN107766320A (en) | A kind of Chinese pronoun resolution method for establishing model and device | |
| CN111737427B (en) | Method for recommending lesson forum posts by combining forum interaction behaviors and user reading preference | |
| CN111709244A (en) | Deep learning method for identifying causal relationship of contradictory dispute events | |
| CN106980650A (en) | A kind of emotion enhancing word insertion learning method towards Twitter opinion classifications | |
| CN108280218A (en) | A kind of flow system based on retrieval and production mixing question and answer | |
| CN113011196B (en) | Concept-enhanced representation and one-way attention-containing subjective question automatic scoring neural network model | |
| CN112232087A (en) | Transformer-based specific aspect emotion analysis method of multi-granularity attention model | |
| CN110489554B (en) | Attribute-level emotion classification method based on location-aware mutual attention network model | |
| CN110059220A (en) | A kind of film recommended method based on deep learning Yu Bayesian probability matrix decomposition | |
| CN115510814B (en) | Chapter-level complex problem generation method based on dual planning | |
| CN113361269B (en) | Text emotion classification method |
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 |