Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can be using the method for pushed information of the application or the implementation of the device for pushed information
The exemplary architecture 100 of example.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 to provide the medium of communication link.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
Terminal device 101,102,103 is interacted by network 104 with server 105, to receive or send message etc..Terminal
Various client applications can be installed in equipment 101,102,103.For example, short video class is applied, web browser class is applied,
Picture shares class application, news category application etc..
Terminal device 101,102,103 can be hardware, be also possible to software.When terminal device 101,102,103 is hard
When part, it can be the various electronic equipments for supporting information push and voice input, including but not limited to smart phone, plate electricity
Brain, E-book reader, pocket computer on knee and desktop computer etc..When terminal device 101,102,103 is software
When, it may be mounted in above-mentioned cited electronic equipment.Multiple softwares or software module may be implemented into (such as mentioning in it
For the multiple softwares or software module of Distributed Services), single software or software module also may be implemented into.It does not do herein specific
It limits.
Server 105 can be to provide the server of various services, such as support to be pacified on terminal device 101,102,103
The back-end server of the client application of dress.Back-end server can receive client transmission, for requesting pushed information
Voice data, and handled by the analysis etc. to voice data, determine information to be pushed.It later, can be by determining wait push
Information is back to terminal device.
It should be noted that the method provided by the embodiment of the present application for pushed information is generally held by server 105
Row, correspondingly, the device for pushed information is generally positioned in server 105.
It should be noted that server can be hardware, it is also possible to software.When server is hardware, may be implemented
At the distributed server cluster that multiple servers form, individual server also may be implemented into.It, can when server is software
It, can also be with to be implemented as multiple softwares or software module (such as providing multiple softwares of Distributed Services or software module)
It is implemented as single software or software module.It is not specifically limited herein.
It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.
With continued reference to Fig. 2, it illustrates the processes according to one embodiment of the method for pushed information of the application
200.This for pushed information method the following steps are included:
Step 201, voice data that client is sent, for requesting pushed information is received.
In the present embodiment, it can be connect for the executing subject of the method for pushed information (server 105 as shown in Figure 1)
Voice data that receipts client (terminal device 101,102,103 as shown in Figure 1) are sent, for requesting pushed information.
User can be used to request pushed information when wanting to obtain pushed information by the client typing that it is used
Voice data.The voice data for being used to request pushed information of user's typing can be sent to above-mentioned executing subject by client.
Step 202, voice data is input to disaggregated model trained in advance, obtains sorting result information.
In the present embodiment, sorting result information can be used to indicate that pre- belonging to the requested pushed information of voice data
If classification.It specifically, in advance can be the requested pushed information of voice data according to actual application demand by technical staff
One or more classifications are set.Wherein, disaggregated model can classify to the voice data received, to determine voice data
Requested pushed information belongs to which preset classification.
It is exactly the real demand for needing to analyze user to a main problem of user's pushed information, identification is used in other words
The intention at family.A kind of common mode is namely based on the historical behavior data of user, predicts the real demand of user.Due to dividing
It is precipitated after the real demand of user, the range of search of information to be pushed can be reduced, therefore can retrieve higher-quality
Information, and pushed to user.
Based on this, in the present embodiment, by technical staff in advance by the real demand of user in other words user intention draw
It is divided into several classifications, when receiving user, voice data for requesting pushed information, belonging to the intention that first determines user
Pre-set categories.It is then based on pre-set categories and determines information to be pushed, more fit in the true of user so as to can push to user
The information of real demand.
It should be appreciated which classification be specifically divided into, and several classifications of division etc. can be according to actual
Business demand and product demand determine.For example, one is used dedicated for the server of push music category information and one are special
The classification type set by the server of push finance and economics category information, which may have, to be very different.
Optionally, above-mentioned disaggregated model can be trained as follows and be obtained:
Step 1 obtains training sample set.Wherein, the training sample that training sample is concentrated may include for requesting push
The voice data of information, and the classification results letter for indicating pre-set categories belonging to the requested pushed information of voice data
Breath.
In practice, the corresponding sorting result information of voice data in each training sample can be marked by technical staff.
Step 2 determines preliminary classification model.Wherein, preliminary classification model can be it is various types of unbred or
The not artificial neural network that training is completed.Such as preliminary classification model can be deep learning model.Preliminary classification model can also
To be the model being combined to artificial neural network a variety of unbred or that training is not completed.For example, initial point
Class model can be to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and unbred full connection
The model that layer is combined.
The specific network structure of preliminary classification model (such as needs to include which layer, every layer of the number of plies, the size of convolution kernel
Deng) can be determined by technical staff according to actual application demand.
Step 3, using the method for machine learning, by the language in the training sample in the training sample of training sample concentration
Input of the sound data as preliminary classification model, sorting result information corresponding with the voice data of input is defeated as it is expected
Out, training obtains above-mentioned disaggregated model.
Specifically, can the loss function (such as cross entropy) based on model common, suitable for training for classification come
Training preliminary classification model.Wherein, the value of loss function may be used to determine the reality output and expectation of model in training process
Output degree of closeness.It is then possible to which the value based on loss function, adjusts preliminary classification model using the method for backpropagation
Parameter terminate training and in the case where meeting preset trained termination condition.After the completion of training, training can be completed
Preliminary classification model be determined as above-mentioned disaggregated model.
Wherein, preset trained termination condition can include but is not limited at least one of following: the training time is more than default
Duration, frequency of training are met certain condition more than preset times, the value of loss function.
In some optionally implementations of the present embodiment, disaggregated model may include that text determines model and text point
Class model.At this point it is possible to which voice data, which is first input to text, determines model, the corresponding text data of voice data is obtained.So
Afterwards, text data can be input to textual classification model, obtains sorting result information.
Wherein, text determines that model can be obtained based on existing various audio recognition methods.It is alternatively possible to directly obtain
With taking away source, trained speech recognition modeling as text determines model.Alternatively it is also possible to using existing some languages
The net of sound identification model (Hidden Markov Model such as based on parameter model, the Vector Quantization algorithm based on nonparametric model)
Network structure obtains text using training data training and determines model.
Textual classification model can be trained as follows and be obtained:
Step 1 obtains training sample set.Wherein, the training sample that training sample is concentrated may include for requesting push
The corresponding text data of the voice data of information, and for indicating default class belonging to the requested pushed information of voice data
Other sorting result information.
In practice, the corresponding sorting result information of text data in each training sample can be marked by technical staff.
Step 2 determines original text disaggregated model.Wherein, original text disaggregated model can be by technical staff according to reality
The application demand (such as needing to include which layer, every layer of the number of plies, the size of convolution kernel) on border is constructed, and can also be chosen existing
The some training having are completed or unfinished textual classification model (such as Text-CNN (text convolutional neural networks), Text-RNN
(text Recognition with Recurrent Neural Network) etc.) it is used as original text disaggregated model.
Step 3, using the method for machine learning, by the text in the training sample in the training sample of training sample concentration
Input of the notebook data as original text disaggregated model, using sorting result information corresponding with the text data of input as expectation
Output, training obtain above-mentioned textual classification model.
Specifically, can the loss function (such as cross entropy) based on model common, suitable for training for classification come
The above-mentioned original text disaggregated model of training.Wherein, the value of loss function may be used to determine model in training process reality it is defeated
Out with the degree of closeness of desired output.It is then possible to which the value based on loss function, is adjusted initial using the method for backpropagation
The parameter of textual classification model, and in the case where meeting preset trained termination condition, terminate training.It, can after the completion of training
The original text disaggregated model of completion will be trained to be determined as above-mentioned textual classification model.
Wherein, preset trained termination condition can include but is not limited at least one of following: the training time is more than default
Duration, frequency of training are met certain condition more than preset times, the value of loss function.
Step 203, according to sorting result information, information to be pushed is determined.
In the present embodiment, can using sorting result information as the one aspect that considers when determining information to be pushed, with
The corresponding user of client for allowing identified information to be pushed to meet transmission voice data as much as possible really wants
Pushed information.The specific method for determining information to be pushed can flexibly change.
As an example, candidate information to be pushed collection first can be determined according to the historical behavior data of user.Then from candidate
Information to be pushed concentrates the candidate information to be pushed for choosing the pre-set categories for belonging to sorting result information instruction as letter to be pushed
Breath.Similarly, naturally it is also possible to the first matched candidate information to be pushed collection of the determining pre-set categories indicated with sorting result information,
Then it is concentrated from candidate information to be pushed, according to the historical behavior data of user, chooses candidate pushed information and be used as letter to be pushed
Breath.
As an example, can also be preset different according to the attribute of the corresponding information to be pushed of various pre-set categories
Determine the algorithm of information to be pushed.At this point, after obtaining sorting result information, it can be to voice data execution and classification results
The pre-set categories corresponding algorithm of information instruction determines information to be pushed.
In practice, corresponding algorithm can be executed to determine wait push by above-mentioned executing subject according to sorting result information
Information.Alternatively, the terminal device for corresponding respectively to various pre-set categories can also be preset, respectively according to various default classes
Other voice data, determines information to be pushed.At this point, above-mentioned executing subject is after obtaining sorting result information, it can be by language
Sound data are sent to terminal device corresponding with the classification of sorting result information instruction.Letter to be pushed is determined by terminal device
It ceases and sends and supreme state executing subject.
Step 204, to client push information to be pushed.
The method provided by the above embodiment of the application, which is realized using voice data, triggers push information to be pushed
Interactive mode.And to what is received be used to that the voice data of pushed information to be requested to be classified, according to belonging to voice data
Classification determine information to be pushed, to help so that the information to be pushed determined can be closer in voice data pair
The true intention of the user answered, while user can also be saved and obtain its required information the time it takes, required flower
Interaction cost taken etc..
With further reference to Fig. 3, it illustrates the processes 300 of another embodiment of the method for pushed information.The use
In the process 300 of the method for pushed information, comprising the following steps:
Step 301, voice data that client is sent, for requesting pushed information is received.
Step 302, voice data is input to disaggregated model trained in advance, obtains sorting result information.
The specific implementation procedure of above-mentioned steps 301 and 302 can refer to step 201 in Fig. 2 corresponding embodiment and 202
Related description, details are not described herein.
Step 303, in response to determining that sorting result information indicates to preset belonging to the requested pushed information of voice data
Classification is the first pre-set categories, determines the subcategory information of the requested pushed information of voice data;From with subcategory information pair
The information answered, which is concentrated, chooses information as information to be pushed.
In this step, subcategory information can be used to indicate that belonging to the requested pushed information of voice data, preset
Default subclass in classification.First pre-set categories can be used to indicate that the requested pushed information of voice data can also be into one
Step is divided into a thinner default subclass.In other words, the information to be pushed that user wants is to belong to a certain default subclass
Other information.
In this step, multiple default subclass can be preset according to actual application scenarios by technical staff.Example
Such as, if above-mentioned executing subject be used for the various pictures of client push, then can preset: landscape, personage, animation, in
The multiple subclass of national practice of forms of behavior etc..If above-mentioned executing subject is used for the various videos of client push, then can preset: joy
The multiple subclass of pleasure, sport, finance and economics etc..
In this step, determining that the information to be pushed that the corresponding user of voice data wants is to belong to a certain default subclass
After other information, the subcategory information of the requested pushed information of voice data can be determined using various methods.
It is alternatively possible to the class model of training subdivision in advance, and determine that voice data is requested using subdivision class model and push away
It delivers letters the subcategory information of breath.Specifically, voice data can be input to subdivision class model trained in advance, obtain voice number
According to the subcategory information of requested pushed information.
Wherein, subdivision class model can train as follows obtains:
Step 1 obtains training sample set.Wherein, the training sample that training sample is concentrated may include for requesting push
The subcategory information of the requested pushed information of voice data and voice data of information.
In practice, the corresponding subcategory information of voice data in each training sample can be marked by technical staff.
Step 2 determines initial subdivision class model.Wherein, initially subdivision class model can be various types of indisciplines
Or not training complete artificial neural network.Such as initial subdivision class model can be deep learning model.Initial disaggregated classification
Model is also possible to the model being combined to artificial neural network a variety of unbred or that training is not completed.Example
Such as, initially subdivision class model can be to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and without
The model that trained full articulamentum is combined.
The specific network structure of initial subdivision class model (such as needs to include which layer, every layer of the number of plies, the size of convolution kernel
Deng) can be arranged by technical staff according to actual application demand.
Step 3, using the method for machine learning, by the language in the training sample in the training sample of training sample concentration
Input of the sound data as initial subdivision class model, using the subcategory information of the voice data of input as desired output, training
Obtain above-mentioned subdivision class model.
Specifically, can the loss function (such as cross entropy) based on model common, suitable for training for classification come
The initial subdivision class model of training.Wherein, the value of loss function may be used to determine the reality output of model and phase in training process
The degree of closeness of the output of prestige.It is then possible to which the value based on loss function, adjusts initial disaggregated classification using the method for backpropagation
The parameter of model, and in the case where meeting preset trained termination condition, terminate training.It, can be by training after the completion of training
The initial subdivision class model completed is determined as above-mentioned subdivision class model.
Wherein, preset trained termination condition can include but is not limited at least one of following: the training time is more than default
Duration, frequency of training are met certain condition more than preset times, the value of loss function.
It is alternatively possible to first determine that the corresponding text data of voice data determines voice data then according to text data
Subcategory information.
Wherein it is possible to determine the corresponding text data of voice data using existing various speech recognition algorithms.And it is based on
Text data can also determine the corresponding subcategory information of voice data using a variety of different methods.
It is alternatively possible to be directed to every kind of default subclass, corresponding keywords database is preset.Voice data is being determined
After corresponding text data, the keyword that existing various keyword extraction algorithms extract text data can use.Later,
The similarity of the keyword of the text data of extraction keywords database corresponding with each default subclass can be calculated.Then it chooses
The subcategory information of the corresponding default subclass of maximum similarity is as the corresponding subcategory information of voice data.
Wherein, keywords database corresponding for any default subclass, the keywords database and extracted text data
The similarity of keyword can use the similarity of each keyword in the keyword and the keywords database of the text data extracted
Average value or weighted average indicate.
It is alternatively possible to which text data, which is input to subcategory information trained in advance, determines model, subclass letter is obtained
Breath.Wherein, subcategory information determines that model can train as follows and obtains:
Step 1 obtains training sample set.Wherein, the training sample that training sample is concentrated may include for requesting push
The subcategory information of the corresponding text data of the voice data of information and the requested pushed information of voice data.
In practice, the corresponding subcategory information of text data in each training sample can be marked by technical staff.
Step 2 determines that initial subcategory information determines model.Wherein, initial subcategory information determines that model can be by skill
Art personnel construct according to actual application demand (such as needing to include which layer, every layer of the number of plies, the size of convolution kernel).
Initial subcategory information determines that model can also choose existing some training completions or unfinished textual classification model (such as
Text-CNN (text convolutional neural networks), Text-RNN (text Recognition with Recurrent Neural Network) etc.) it is used as original text disaggregated model.
Optionally, initial subcategory information determines that model may include that initial characteristics extract model and the classification of initial subclass
Model.Wherein, initial characteristics, which extract model, can be used for extracting text data, preset attribute attribute value.Wherein, it presets and belongs to
Property can be arranged by technical staff according to actual application scenarios.As an example, preset attribute includes but is not limited to: textual data
Time, geographical location of text attribute instruction for being indicated according to the personage of instruction, text data etc..It is defeated that initial characteristics extract model
Attribute value out can be used as the input of initial subclass disaggregated model, and initial subclass disaggregated model can export subclass letter
Breath.
Specifically, the more existing model for slot filling (slot filling) can be chosen as above-mentioned initial spy
Sign extracts model, can choose textual classification model (such as Text-CNN (the text volume that existing some training are completed or do not completed
Product neural network), Text-RNN (text Recognition with Recurrent Neural Network) etc.) classify mould as the initial subclass of original text disaggregated model
Type.
Step 3, using the method for machine learning, by the text in the training sample in the training sample of training sample concentration
Notebook data determines the input of model as initial subcategory information, using subcategory information corresponding with the text data of input as
Initial subcategory information determines the desired output of model, and training obtains above-mentioned subcategory information and determines model.
Specifically, can the loss function (such as cross entropy) based on model common, suitable for training for classification come
The above-mentioned initial subcategory information of training determines model.Wherein, the value of loss function may be used to determine model in training process
The degree of closeness of reality output and desired output.It is then possible to the value based on loss function, using the method tune of backpropagation
Whole initial subcategory information determines the parameter of model, and in the case where meeting preset trained termination condition, terminates training.Instruction
After the completion of white silk, the initial subcategory information that training is completed can be determined that model is determined as above-mentioned subcategory information and determines model.
Wherein, preset trained termination condition can include but is not limited at least one of following: the training time is more than default
Duration, frequency of training are met certain condition more than preset times, the value of loss function.
It should be noted that determining what model was made of two or more submodels for initial subcategory information
Situation, such as when initial subcategory information determines that model extracts model and initial subclass disaggregated model by initial characteristics and forms,
One of submodel can also be first trained, the parameter of that trained submodel, another submodule of retraining are then fixed
Type.Training method specifically is varied, and training data needed for different training methods may be different.The application couple
This is with no restrictions.
Step 304, in response to determining that sorting result information indicates to preset belonging to the requested pushed information of voice data
Classification is the second pre-set categories, is retrieved, is retrieved in the corresponding information concentration of the second pre-set categories using voice data
Result set;It is concentrated from search result and chooses search result as information to be pushed.
In this step, the second pre-set categories can be used to indicate that the requested pushed information of voice data is voice data
The answer in other words of the search result of expressed problem.In this case, what voice data indicated can be a problem, that is, use
The answer for the problem of desired information to be pushed in family is asked.
It should be appreciated that voice data can be one for enquirement interrogative sentence, be also possible to certainly clause, but use
In one problem of expression.For example, the corresponding text data of voice data can be " what the piece caudal flexure of XX TV is? ", can also
Think " I wants to listen the piece caudal flexure of XX TV ".
In this step, the corresponding information collection of the second pre-set categories can be preset.Wherein, information collection can be above-mentioned
The information collection that executing subject is stored is also possible to information collection provided by third party's data source.Determining that voice data is asked
Pre-set categories belonging to the pushed information asked be the second pre-set categories after, can based on voice data in the second pre-set categories pair
The information concentration answered is retrieved, to obtain retrieval set.
Later, search result can be chosen from retrieval set as information to be pushed.For example, can be from retrieval set
In randomly select a search result, can also from search result concentrate choose first search result as information to be pushed.
Step 305, in response to determining that sorting result information indicates to preset belonging to the requested pushed information of voice data
Classification is third pre-set categories, concentrates in the corresponding information of third pre-set categories and chooses information as information to be pushed.
In this step, third pre-set categories can be used to indicate that the requested pushed information of voice data can be any
Pushed information, i.e. user thinks at will to browse information, and the content of the information of browsing is wanted without explicitly limiting.
In this case, it can directly be concentrated in the corresponding information of third pre-set categories and choose information as letter to be pushed
Breath.Specifically choosing mode can be with flexible setting.For example, letter can be randomly selected from the corresponding information concentration of third pre-set categories
Breath is used as information to be pushed, the information for choosing newest storage can also be concentrated to be used as wait push away from the corresponding information of third pre-set categories
It delivers letters breath etc..Wherein, the corresponding information collection of third pre-set categories can be preassigned by technical staff.
Step 306, to client push information to be pushed.
With continued reference to the signal that Fig. 4, Fig. 4 are according to the application scenarios of the method for pushed information of the present embodiment
Figure 40 0.In the application scenarios of Fig. 4, user can by using client recording be used to request the voice number of pushed information
According to 401, it is then forwarded to above-mentioned executing subject.As shown in the figure, voice data is " I wants to see the ball match of XX ".Above-mentioned execution master
Voice data can be input to disaggregated model 402 after receiving voice data 401 by body.
As shown in the figure, disaggregated model 402 for determine input voice data belong to preset " the first pre-set categories ",
Any pre-set categories in " the second pre-set categories ", " third pre-set categories ".First pre-set categories can be used to indicate that voice
The requested pushed information of data can be with further division into a thinner default subclass.Second pre-set categories can be with
For indicating that the requested pushed information of voice data is the answer of problem expressed by voice data.Third pre-set categories can be with
For indicating that the requested pushed information of voice data can be arbitrary pushed information.
As shown in figure label 403, voice data 401 belongs to the first pre-set categories.It therefore, can be by voice data 401
It is input to subcategory information and determines model 404.As shown in the figure, subcategory information determine model 404 for determine input language
Sound data belong to which of preset " amusement ", " sport ", " finance and economics " subclass.
As shown in figure label 405, voice data 401 belongs to " sport " this subclass.Therefore, can from " sport " this
Pushed information is chosen in the corresponding information collection 406 of one subclass, i.e. " ball match " 407 needed for user is sent to client.
From figure 3, it can be seen that the method for pushed information compared with the corresponding embodiment of Fig. 2, in the present embodiment
Process 300 highlight the pushed information of user demand be divided into three kinds of pre-set categories.One is users to want arbitrarily browsing
Some information, one is users to propose problem, it is desirable to the answer for obtaining problem, it can be into one is the pushed information needed for user
One step is divided to some default subclass.The difference of pre-set categories belonging to information according to demand, it is corresponding using different
Method determines information to be pushed and pushes to terminal device used by a user.Due to pure according to user's history behavioral data
To user's pushed information, it is easy to appear the situation of user's intention assessment mistake, and this side for being finely divided the intention of user
Formula can allow the received pushed information of user more to meet the real demand of user.In addition, the push needed for user is believed
When breath can preset subclass to some with further division, letter to be pushed can be further chosen under corresponding subclass
Breath, so that user can receive higher-quality pushed information.
With further reference to Fig. 5, as the realization to method shown in above-mentioned each figure, this application provides for pushed information
One embodiment of device, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which specifically can be applied to
In various electronic equipments.
As shown in figure 5, the device 500 provided in this embodiment for pushed information includes receiving unit 501, taxon
502, determination unit 503 and push unit 504.Wherein, receiving unit 501, be configured to receive client transmission, for asking
Seek the voice data of pushed information;Taxon 502 is configured to for voice data being input to disaggregated model trained in advance,
Obtain sorting result information, wherein sorting result information is preset belonging to the requested pushed information of voice data for indicating
Classification;Determination unit 503 is configured to determine information to be pushed according to sorting result information;Push unit 504, is configured to
To client push information to be pushed.
In the present embodiment, in the device of pushed information 500: receiving unit 501, taxon 502, determination unit
503 and push unit 504 specific processing and its brought technical effect can be respectively with reference to the step in Fig. 2 corresponding embodiment
201, the related description of step 202, step 203 and step 204, details are not described herein.
In some optional implementations of the present embodiment, disaggregated model includes that text determines model and text classification mould
Type;And taxon 502 is further configured to: voice data being input to text and determines model, obtains voice data pair
The text data answered;Text data is input to textual classification model, obtains sorting result information.
In some optional implementations of the present embodiment, determination unit 503 is further configured to: in response to determination
Sorting result information indicates that pre-set categories belonging to the requested pushed information of voice data are the first pre-set categories, determines voice
The subcategory information of the requested pushed information of data, wherein subcategory information is for indicating the requested push of voice data
Default subclass belonging to information, in pre-set categories;From information corresponding with subcategory information concentrate choose information be used as to
Pushed information.
In some optional implementations of the present embodiment, determination unit 503 is further configured to: determining voice number
According to corresponding text data;According to text data, subcategory information is determined.
In some optional implementations of the present embodiment, determination unit 503 is further configured to: by text data
It is input to subcategory information trained in advance and determines model, obtain subcategory information.
In some optional implementations of the present embodiment, determination unit 503 is further configured to: in response to determination
Sorting result information indicates that pre-set categories belonging to the requested pushed information of voice data are the second pre-set categories, utilizes voice
Data are retrieved in the corresponding information concentration of the second pre-set categories, obtain retrieval set;It is concentrated from search result and chooses inspection
Hitch fruit is as information to be pushed.
In some optional implementations of the present embodiment, determination unit 503 is further configured to: in response to determination
Sorting result information indicates that pre-set categories belonging to the requested pushed information of voice data are third pre-set categories, pre- in third
If the corresponding information of classification, which is concentrated, chooses information as information to be pushed.
The device provided by the above embodiment of the application, by receiving unit receive client send, for request push away
It delivers letters the voice data of breath;Voice data is input to disaggregated model trained in advance by taxon, obtains sorting result information,
Wherein, sorting result information is for indicating pre-set categories belonging to the requested pushed information of voice data;Determination unit according to
Sorting result information determines information to be pushed;Push unit is to client push information to be pushed, to provide a user voice
The interactive mode of information push is triggered, and by analyzing pre-set categories belonging to the requested pushed information of voice data, to
Client push information corresponding with pre-set categories, realize more refine information push, and then facilitate promoted user's viscosity,
The activity of the user and retention ratio.
Below with reference to Fig. 6, it illustrates the computer systems 600 for the server for being suitable for being used to realize the embodiment of the present application
Structural schematic diagram.Server shown in Fig. 6 is only an example, should not function and use scope band to the embodiment of the present application
Carry out any restrictions.
As shown in fig. 6, computer system 600 includes central processing unit (CPU) 601, it can be read-only according to being stored in
Program in memory (ROM) 602 or be loaded into the program in random access storage device (RAM) 603 from storage section 608 and
Execute various movements appropriate and processing.In RAM 603, also it is stored with system 600 and operates required various programs and data.
CPU 601, ROM 602 and RAM 603 are connected with each other by bus 604.Input/output (I/O) interface 605 is also connected to always
Line 604.
I/O interface 605 is connected to lower component: the importation 606 including keyboard, mouse etc.;It is penetrated including such as cathode
The output par, c 607 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section 608 including hard disk etc.;
And the communications portion 609 of the network interface card including LAN card, modem etc..Communications portion 609 via such as because
The network of spy's net executes communication process.Driver 610 is also connected to I/O interface 605 as needed.Detachable media 611, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 610, in order to read from thereon
Computer program be mounted into storage section 608 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 609, and/or from detachable media
611 are mounted.When the computer program is executed by central processing unit (CPU) 601, limited in execution the present processes
Above-mentioned function.
It should be noted that the computer-readable medium of the application can be computer-readable signal media or computer
Readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but it is unlimited
In system, device or the device of --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or any above combination.It calculates
The more specific example of machine readable storage medium storing program for executing can include but is not limited to: have the electrical connection, portable of one or more conducting wires
Formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable programmable read only memory
(EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device or
The above-mentioned any appropriate combination of person.In this application, computer readable storage medium can be it is any include or storage program
Tangible medium, which can be commanded execution system, device or device use or in connection.And in this Shen
Please in, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
In carry computer-readable program code.The data-signal of this propagation can take various forms, including but not limited to
Electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer-readable
Any computer-readable medium other than storage medium, the computer-readable medium can send, propagate or transmit for by
Instruction execution system, device or device use or program in connection.The journey for including on computer-readable medium
Sequence code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc. are above-mentioned
Any appropriate combination.
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor, packet
Include receiving unit, taxon, determination unit and push unit.Wherein, the title of these units not structure under certain conditions
The restriction of the pairs of unit itself, for example, receiving unit be also described as " receive it is that client is sent, for requesting to push away
Deliver letters breath voice data unit ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in server described in above-described embodiment;It is also possible to individualism, and without in the supplying server.It is above-mentioned
Computer-readable medium carries one or more program, when said one or multiple programs are executed by the server,
So that the server: receiving voice data that client is sent, for requesting pushed information;Voice data is input in advance
Trained disaggregated model, obtains sorting result information, wherein sorting result information is for indicating the requested push of voice data
Pre-set categories belonging to information;According to sorting result information, information to be pushed is determined;To client push information to be pushed.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.