Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, completeSite preparation description, it is clear that described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based onEmbodiment in the application, it is obtained by those of ordinary skill in the art without making creative efforts every otherEmbodiment shall fall in the protection scope of this application.
In recent years, as the acceleration of the technologies such as mobile Internet, big data, cloud computing, sensor is broken through and is widely applied,The development of artificial intelligence also enters a brand-new stage.And intelligent sound technology is as the key on AI industry's chainOne ring, AI (Artificial Intelligence, artificial intelligence) apply most mature one of technology, in marketing customer service, intelligenceThe fields such as household, intelligent vehicle-carried, intelligence wearing suffer from fast development.For example, intelligent customer service robot.
With the development of internet, social media, in addition to industry traditionally in great demand to customer service (such as operatorCustomer service, bank's customer service, government department's policy answer online robot etc.), some new industries such as: mobile phone, automobile, express deliveryIndustry etc. also all begins trying to introduce virtual customer service assistant (i.e. intelligent customer service robot).For example, as a kind of mode, Yong HukeText or voice messaging are inputted in human-computer interaction input interface with terminal (such as mobile phone), intelligent customer service robot is to inputText identified, or the text being converted to by the voice inputted is identified, then by corresponding answer text withThe mode of text or voice feeds back to user, to realize intelligent interaction.
Inventor has found under study for action, can be spoken by allowing virtual customer service assistant to simulate true man, the cordiality of Lai Zengjin customer serviceSense.Such as when virtual customer service assistant and user engage in the dialogue, the reply content seeked advice from user can be passed through into virtual peopleThe mouth of figure image, is expressed by way of voice, and user, which can intuitively see on human-computer interaction interface, has voidIntend figural virtual customer service assistant " speaking ", makes the ditch for being able to carry out " face-to-face " between user and virtual customer service assistantLogical exchange.
However, in actual research process, inventors have found that since visual perception of the people to face is more sensitive,In virtual customer service assistant " speaking ", if the virtual image of virtual customer service assistant is constant always, as user and virtual customer service helpManage the difference of interacted content, virtual image when virtual customer service assistant replys the question and answer of user unavoidably can be to user with next lifeHard and extremely unnatural feeling, influences user experience.
In order to improve the above problem, inventor has studied difficulty of the customer service robot to personalize during realization,Even more comprehensively consider the use demand in practical interaction scenarios, proposes the matched side of intelligent customer service in the embodiment of the present applicationMethod, device, electronic equipment and storage medium.
For convenient for better understand the matched method, apparatus of intelligent customer service provided by the embodiments of the present application, terminal device andStorage medium is below first described the application environment for being suitable for the embodiment of the present application.
Referring to Fig. 1, Fig. 1 shows a kind of application environment schematic diagram suitable for the embodiment of the present application.The application is implementedThe matched method of intelligent customer service that example provides can be applied to polymorphic interactive system 100 as shown in Figure 1.Polymorphic interactive system100 include terminal device 101 and server 102, and server 102 and terminal device 101 communicate to connect.Wherein, server 102It can be traditional server, be also possible to cloud server, be not specifically limited herein.
Wherein, terminal device 101 can be with display screen and support the various electronic equipments of data input, including but notIt is limited to smart phone, tablet computer, pocket computer on knee, desktop computer and wearable electronic equipment etc..Specifically,Data input can be based on voice module input voice, the character input module input character etc. having on terminal device 101.
Wherein, client application can be installed, user can be based on client application on terminal device 101(such as APP, wechat small routine etc.) is communicated with server 102.Specifically, being equipped with corresponding service on server 102Application program is held, user can register a user account number in server 102 based on client application, and be based on the userAccount number is communicated with server 102, such as user is in client application login user account number, and is based on the user account numberIt is inputted by client application, can be with inputting word information or voice messaging etc., client application receivesAfter the information of user's input, server 102 can be sent this information to, so that server 102 can receive the information and go forward side by sideRow processing and storage, server 102 can also receive the information and return to a corresponding output information to end according to the informationEnd equipment 101.
In some embodiments, client application can be used for providing a user customer service, carry out with userCustomer service is linked up, and client application can be interacted based on virtual robot with user.Specifically, client applicationIt can receive the information of user's input, and response made to the information based on virtual robot.Wherein, virtual robot is to be based onThe software program of visualized graphs, the software program can show the machine of simulation biobehavioral or thought to user after being performedThe humanoid state of device.Virtual robot can be the robot of simulation true man's formula, such as be built according to user itself or other people formThe robot of vertical likeness in form true man, is also possible to the robot of animation effect formula, such as zoomorphism or cartoon figure's formRobot is not limited thereto.
In some embodiments, terminal device 101, can after obtaining return information corresponding with the information that user inputsTo show the corresponding void with the return information on the display screen of terminal device 101 or other image output devices connected to it(wherein, which may include the gender of virtual robot to quasi- robot graphics, audio institute is right with replyingThe answer emotion answered and image characteristics etc.).It can pass through while playing virtual robot image as a kind of modeThe loudspeaker of terminal device 101 or other audio output apparatus connected to it play sound corresponding with virtual robot imageFrequently, can also in the display screen display text corresponding with the return information or figure of terminal device 101, realize image,In many aspects such as voice, text with the polymorphic interaction of user.
In some embodiments, the device handled the information of user's input also can be set in terminal deviceOn 101, so that terminal device 101 communicates the interaction that can be realized with user with the foundation of server 102 without relying on, it is polymorphic at this timeInteractive system 100 can only include terminal device 101.
Above-mentioned application environment is only for convenience of example made by understanding, it is to be understood that the embodiment of the present application not only officeIt is limited to above-mentioned application environment.
It below will be by specific embodiment to the matched method, apparatus of intelligent customer service provided by the embodiments of the present application, terminalEquipment and storage medium are described in detail.
As shown in Fig. 2, showing the method flow diagram of the matched method of intelligent customer service of one embodiment of the application offer.ThisThe matched method of intelligent customer service that embodiment provides can be applied to that there is the terminal of display screen or other image output devices to setStandby, terminal device can be the electronic equipments such as smart phone, tablet computer, wearable intelligent terminal.
In the particular embodiment, it is matched to can be applied to intelligent customer service as shown in Figure 7 for the matched method of intelligent customer serviceDevice 600 and terminal device shown in Fig. 8 101.It will be explained in detail below for process shown in Fig. 2.Above-mentionedThe matched method of intelligent customer service specifically may comprise steps of:
Step S110: the voice messaging of user in interactive process is obtained.
Wherein, the voice messaging in the embodiment of the present application includes the audio-frequency information of user.For example, it may be user is man-machineThe voice data (i.e. audio-frequency information) that interactive interface is inputted by the speech voice input function of terminal device, is also possible to the language of userSound recording data, or perhaps the voice data of the personage in one section of video, again or can be simulation personage voice dataDeng being not limited thereto.As a kind of mode, the voice data may include user language (such as may determine that be ChineseIt is also non-Chinese, if being Chinese, judgement is mandarin or non-generic words, if non-generic words, which kind of side judgement belongs toSpeech, for example, judgement is Sichuan words or Hunan words etc.), tone color, tone, volume etc., do not enumerate herein.
As a kind of mode, the voice messaging of the user in interactive process can be the current man-machine friendship for passing through terminal deviceThe voice messaging for the user that mutual interface and terminal device interact.It is understood that when user passes through human-computer interaction interfaceWhen interacting with terminal device, ambient enviroment can inevitably be mingled with background sound (including noise, reverberation and echoDeng), interaction effect may be will affect.So, as a kind of mode, in order to promoted user by voice messaging and terminal device intoInteraction effect when row interaction, the voice messaging of user in available interactive process, after being carried out to the voice messagingInteraction effect is promoted after continuous processing, will be specifically illustrated subsequent.
A kind of voice signal as mode, in the voice messaging of available user.It optionally, can be by voiceInformation carries out feature extraction, and then identifies the acoustic model and language model of voice messaging, then to obtained acoustic modelAnd language model is decoded, and obtains text information corresponding with voice messaging.
Optionally, it for the voice messaging of user acquired in terminal device, can store to local, it can also be by terminalEquipment is sent to server and is stored, can be to avoid the storing data of terminal device in such a way that server is storedRedundancy bring arithmetic speed slows down.
Step S120: the voice messaging is inputted into the first nerves network model that training obtains in advance, identifies correspondenceUser personal identification characteristic information.
Wherein, first nerves network model can be RNN (Recurrent Neural Network, Recognition with Recurrent Neural Network)Or the neural network models such as LSTM (Long Short-Term Memory, shot and long term memory network).Personal identification feature letterBreath characterization speaks the characteristic information of identity of user optionally can be by the voice messaging of user to the identity characteristic of userDetermined.For example, it is assumed that user has said sentence " what this noon eats ", then can be according to the voice for the user for saying thisThe approximate age of the gender (e.g. male either women), user that identify user is (for example, the sound of child, youngSound, the either old sound of the sound in middle age etc.), user use when speaking language (e.g. English or Chinese,Mandarin or dialect etc.) and user this moment mood (for example, with laughter sound in other words indignation soundDeng) etc..
The personal identification feature largely about user can be stored in advance as a kind of mode, in first nerves network modelThe sample database of information.Optionally, which may include the training sample of the characteristic information of the identity of a large amount of characterization users.MakeFor a kind of implementation, the first nerves network that training obtains in advance can be inputted by the voice messaging for the user that will acquireModel, to identify the personal identification characteristic information of user corresponding with the voice messaging.
For example, in a specific application scenarios, if first nerves network model is LSTM neural network model, thenAfter voice messaging is input to LSTM neural network model, LSTM neural network model can carry out noise to voice dataFilter out and filter out according to time series the personal identification characteristic information of corresponding user.
Step S130: corresponding Virtual Intelligent customer service is matched to the voice messaging according to the personal identification characteristic informationCarry out response.
It is understood that the voice messaging that different user is issued has differences, difference including language and speakDifference of sound etc., for example, for the elderly that will not be spoken Mandarin, if robot customer service still uses mandarinIt is linked up with the elderly, may there is communication disorder；And for the foreigner that will not be spoken Chinese for one, if robot customer serviceIt is linked up with Chinese with foreigner, it is abnormal to will also result in communication.So, in order to meet the communication need of different crowd Yu robot clientIt asks, the embodiment of the present application can match corresponding Virtual Intelligent customer service according to the personal identification characteristic information of user and believe voiceBreath carries out response.
Wherein, Virtual Intelligent customer service can be understood as that true man can be simulated engages in the dialogue with user, and will be to user'sThe virtual robot that answer content is expressed by way of voice based on figure image makes user can see that robot visitorClothes are being spoken, i.e. robot customer service can be engaged in the dialogue in a manner of visual speech with user.
Optionally, the corresponding Virtual Intelligent customer service that the personal identification characteristic information of different users matches is notTogether.For example, if the voice messaging " I can go kiteflying " according to user identifies that the user that speaks is child, then matchingVirtual Intelligent customer service corresponding with the child can be grow compared with have the character image of affinity, and match correspondingReply audio；If the voice messaging " we go kiteflying " according to user identifies that the user that speaks is male user, then canIt can be the facial image of the sweet women of appearance with matched Virtual Intelligent customer service corresponding with the male；If according toThe voice messaging " this weekend comes my room and plays " at family identifies the language for the user that speaks for Sichuan words, then the matched and user coupleThe Virtual Intelligent customer service answered can be the robot customer service for saying Sichuan.
It optionally, can be by corresponding after matching corresponding Virtual Intelligent customer service according to personal identification characteristic informationVirtual Intelligent customer service carries out response to the voice messaging of user, in order to realize better human-computer interaction, promotes user experience.
Step S140: the Virtual Intelligent customer service is shown.
As a kind of mode, virtual customer service can be shown in the display interface of terminal device, in order to which user can be with voidMore natural human-computer interaction is realized in the communication of quasi- intelligent customer service " face-to-face ".For example, as shown in figure 3, in terminal device 101Display interface show virtual customer service 102, wherein it should be noted that the image of virtual customer service herein is only as one kindExample, the image of virtual customer service can be diversified in actual implementation.As a kind of mode, when user opens video customer serviceWhen function button, Virtual Intelligent customer service can be shown in the user terminal of video customer service.Optionally, for showing Virtual Intelligent customer servicePlace can be unrestricted, for example, it may be the display interface in the APP client of mobile phone is shown, be also possible toIt is shown on the page of the website of operator, is also possible to be shown in the display interface of the terminal devices such as the service machine of bankShow, is specifically not construed as limiting.
The matched method of intelligent customer service provided in this embodiment, by obtaining the voice messaging of user in interactive process, afterAnd voice messaging is inputted into the first nerves network model that training obtains in advance, identify the personal identification feature of corresponding userInformation matches corresponding Virtual Intelligent customer service further according to personal identification characteristic information and carries out response to voice messaging, then showsVirtual Intelligent customer service.It is realized through the above way in the case where obtaining the voice messaging of user, is known according to voice messagingNot Chu user personal identification characteristic information, and then corresponding Virtual Intelligent customer service can be matched according to personal identification characteristic informationIt is shown and response is carried out with the voice messaging to user, so that the interactive mode of user and robot is more true to nature, promote peopleThe vividness and interest of machine interaction, and then enhance user experience.
As shown in figure 4, the method flow diagram for the matched method of intelligent customer service that another embodiment of the application provides is shown,The described method includes:
Step S210: the voice messaging of user in interactive process is obtained.
Step S220: training sample set is obtained.
Wherein, training sample in the embodiment of the present application is concentrated include the training audio extracted from training voice data withAnd personal identification characteristic information corresponding with training audio.Optionally, personal identification characteristic information at least may include user'sGender, age, accent, nationality, emotion are (here it can be appreciated that the mood of user, for example, user says when in a good humorCome if sound with feel blue when say if sound can be different) etc. be used for identity user identityClassification information.
It is understood that in order to preferably match Virtual Intelligent customer service corresponding from different clients, and then promotionThe accuracy matched, the embodiment of the present application can obtain the audio data of several users in advance, and wherein these several users include notSame age bracket, different nationalities, the user of the either same national different regions of different sexes.For example, the voice data obtained canTo include " voice data of tender child, the voice data of sweet little girl, the voice data of cheerful and light-hearted student, change throatThe voice data of the boy of sound, the voice data of clear and melodious melodious music teachers, the magnetic women of tool voice data, alwaysVoice data and the voice data of the foreigner of year people etc. ".Optionally, using the audio data of above-mentioned several users as instructionPractice sample set, so as to enrich audio source and the classification of sample database, so that the result trained is more accurate.
It, can be by including that the terminal device of voice acquisition module acquires the audio number of different user as a kind of modeAccording to for example, enrolling the voice data for obtaining different user with sound pick-up outfit.And alternatively, specific calculation can be passed throughMethod crawls the voice data of different user from network.For example, dissimilarity is obtained from internet by the way of web crawlersNot, different age group, different nationality, different language user voice data, to realize the training sample for obtaining voice dataThis collection.
Step S230: the training sample set is input to machine learning model, and the machine learning model is carried outTraining, obtains first nerves network model.
Optionally, training sample set obtained above is input to machine learning model, and machine learning model is carried outTraining, available first nerves network model.Wherein, machine learning model can be linear model, kernel method and support toAmount machine, decision tree and Boosting and neural network (including full Connection Neural Network, convolutional neural networks, circulation nerve netNetwork etc.) etc..It wherein, can be former with reference to respective work in the prior art about the specific training method of each machine learning modelReason, which is not described herein again.
As a kind of mode, by the way that training sample set is input to machine learning model, available and training sample setIn the corresponding sample database of each sample can be by first nerves net in order to when subsequent any one voice data of inputNetwork model quickly matches the personal identification characteristic information of user corresponding with voice messaging.
Step S240: inputting the first nerves network model for the voice messaging, identifies of corresponding userPeople's identity characteristic information.
Wherein, it is referred to the description in previous embodiment about the particular content of the step, details are not described herein.
Step S250: obtaining answer gender corresponding with the personal identification characteristic information and replies audio frequency characteristics.
Optionally, replying audio frequency characteristics at least may include replying the contents such as used voice, intonation, language.Specifically, as a kind of mode, can be identified by preset answer audio characteristic model corresponding with current speech dataThe corresponding answer gender of the personal identification characteristic information of user and answer audio frequency characteristics.Specifically, in one implementation,Personal identification characteristic information can be input to the nervus opticus network model that training obtains in advance, obtain corresponding answer genderAnd reply audio frequency characteristics.It should be noted that the content trained in nervus opticus network model herein and above-mentioned theThe content trained in one neural network model is different, thereby may be ensured that the accuracy of identification.
Wherein, as a kind of implementation, for different personal identification characteristic informations, corresponding answer gender can be withIt is random, or is matched according to specific rule.For example, for a spadger, available answer genderAlso it is spadger, makes interaction more smooth in order to make spadger see same companion and be unlikely to very contained；And forOne middle-aged male, reply gender can be young woman, in order to promoted middle-aged male interaction enthusiasm so that man-machineInteraction is more abundant, lets user experiencing machine bring close friend experience.
Optionally, for replying audio frequency characteristics, for example, content and use what language time that robot customer service to be answeredIt answers, it can (including language, (category of language including country variant and each country be or not library according to preset rulesWith the dialect type etc. in area) knowledge base and response text library corresponding with types of applications be determined.Wherein, the responseMay include in text library common response text and according to the response combined by machine client intelligent of user voice dataLexicon, so that robot customer service can root when the response text of the voice data of user does not exist in response text libraryChosen inside response lexicon according to the voice data of user the relevant word combination of vocabulary in the voice data to user atResponse text carries out response with the voice data for carrying out to user).
It is understood that the corresponding answer gender of different personal identification characteristic informations and reply audio frequency characteristics can be withDifference is experienced so as to the intelligent user of hoisting machine people's customer service.
It is understood that the same user may be with the variation of mood during interacting with robot customer serviceAnd different language is said, for example, the user that speaks several dialects one while can agree several languages in other words simultaneouslyThe user of speech, then a voice-over actor can imitate various sound.So, in order to promoted human-computer interaction interest and multiplicityProperty, it can be with the voice messaging of real-time detection user with the presence or absence of variation, i.e., the voice data for the user that currently speaks is with the presence or absence of changeChange, optionally, and if it exists, personal identification characteristic information is so input to the nervus opticus network model that training obtains in advance,The available answer gender being adapted to voice variation characteristic and answer audio frequency characteristics.For example, a user start beDialect can be this and adopt with terminal device interaction then the answer gender being adapted to the dialect can be women and answer audioWith the answer audio of dialect, and if the user then uses mandarin, then the answer gender being adapted to skin call canTo be male and reply audio and can be answer audio using mandarin, and the user find the gender of robot customer service withAnd in the case that answer audio can change, in order to meet curiosity, and interacted using English with robot customer service, thenThe answer gender of robot customer service can be adapted for girl and reply audio to be the answer audio using English, thus greatlyThe flexibility and intelligence of robot customer service are enhanced, user experience is promoted.
Step S260: answer affective characteristics corresponding with the personal identification characteristic information are obtained.
It is understood that under informal occasion, the mood of user usually may be used for most of userTo judge from the sound of speaking of user, such as one has just been disturbed by making noise the sound of a word that the user of frame said with othersIt is that there may be more apparent differences that the user of sound and a first place of just having competed, which says this with the sound of a word,'s；The sound of a word that a user says with smile out in other words and a user to speak while crying say same oneThe sound of words also can be different.So, it as a kind of mode, can be obtained from the training sample database of audio data in advanceGet the mapping table of the sound of the user of different mood classifications, stored in the mapping table different moods with it is correspondingSound, in order to after obtaining the voice data of a certain user, be searched from the mapping table and the voice dataCorresponding answer affective characteristics.Wherein, answer affective characteristics here can be understood as robot client and think user feedback responseExpression, mood of robot customer service when information either expression in the eyes etc. can indicate the feature of emotion.
As a kind of mode, personal identification characteristic information can be input to the third nerve network mould that training obtains in advanceType obtains corresponding answer affective characteristics.Wherein, third nerve network model is different from nervus opticus network model.Specifically,Training content in third nerve network is different from the training content in nervus opticus network, so that believing by personal identification featureWhen the breath input third nerve network model that training obtains in advance, the personal identification with active user can be accurately obtainedThe corresponding answer affective characteristics of characteristic information, while can also avoid misidentifying to a certain extent.
Optionally, in order to promote interactive validity, can collect in advance some can guide user or bring userThe mood of some encouragement and positive energy.For example, robot customer service is in response for the voice data for the anxious user that speaksIt can go to help to change user with a kind of gentle, deeply concerned tone, friendly expression of smiling, so that user no longer so worries, butIt adjusts mentality, cools down and solve the problems, such as, optionally, the answer affective characteristics of specific robot customer service in the embodiment of the present applicationCan be unrestricted, it may be any type of answer emotion, for example, facial expression assists answering with the gesture of robot customer serviceMultiple emotion etc..
Step S270: Virtual Intelligent customer service image characteristics corresponding with the personal identification characteristic information are obtained.
Wherein, the image characteristics of Virtual Intelligent customer service can be understood as the features such as figure, bearing, the dress of virtual customer service.As a kind of mode, various types of Virtual Intelligent customer services images can be constructed in advance, and optionally, building mode here can be withIt is to be constructed according to figure image classification, e.g. building personage draws a portrait, and relevant customer service is vivid still to construct cartoon figure's customer serviceImage, for example, the customer service image of building women and the customer service image of building male, the customer service image of building cartoon character or structureBuild the customer service image of pet image.
Optionally, it is to be understood that for different places, different classes of customer service image can be constructed.For example, rightIn bank's enquiry machine system, then the customer service image of personage's portrait can be constructed, for the inquiry system of cinema, can constructThe virtual customer service image of cartoon character, and for places such as zoos, the virtual customer service image of pet image can be constructed.It is differentPlace, so that the image of virtual customer service is more abundant with diversification, user can be promoted using different virtual customer service imagesEntertaining experience.Optionally, it can also be according to age range section and construct, such as building " child, youth, middle age, old age "Etc. classifications section Virtual Intelligent customer service image, wherein may include a variety of customer services images, example under the age range of each classificationSuch as, age range section is that the virtual customer service image of " child " can be according to the virtual customer service shape for the child that the age sorts from small to largeAs being also possible to a variety of virtual customer service images being arranged successively according to the style of wearing the clothes of child, being also possible to the length according to childThe customer service image of plurality of classes that phase character is divided (for example, plump, lovely, sweet, modest, cruel) is specialSign.The building mode of specific Virtual Intelligent customer service image is herein not as restriction.
It should be noted that above-mentioned acquisition answer gender corresponding with personal identification characteristic information and answer audio are specialSign, the acquisition sequence for the image characteristics for replying affective characteristics and Virtual Intelligent customer service can be adjusted according to the actual situation,And acquired content is not as restriction, for example, available more or less feature related with Virtual Intelligent customer service,For example, obtain answer gender corresponding with personal identification characteristic information and reply audio frequency characteristics, reply affective characteristics, withAnd on the basis of the image characteristics of Virtual Intelligent customer service, gesture classification, the gesture class of Virtual Intelligent customer service can be obtained againDeng not illustrating one by one herein.
It is worth noting that, obtaining the mode this embodiments of the image characteristics of Virtual Intelligent customer service not as limit can also be withIt is to be constructed by way of three-dimensional reconstruction, for example, the personally three-dimensional mould of characteristic information setting Virtual Intelligent customer serviceType, then reconstruction model, the principle for being specifically referred to three-dimensional reconstruction realize process, and details are not described herein.
Step S280: audio frequency characteristics, answer emotion spy by the answer gender and are replied by preset algorithmSign and the Virtual Intelligent customer service image characteristics are integrated, and are generated Virtual Intelligent customer service and answered the voice messagingIt answers.
Wherein, preset algorithm can be it is pre-stored can be by the answer gender of above-mentioned acquisition and to reply audio specialSign, answer affective characteristics and Virtual Intelligent customer service image characteristics carry out integrated algorithm, optionally, for example, it may be a set ofEfficient program frame, by that will reply gender and reply audio frequency characteristics, answer affective characteristics and Virtual Intelligent customer service shapeAs feature input is entered, it is vivid to input to obtain a complete Virtual Intelligent customer service, in order to pass through the void of generationQuasi- intelligent customer service carries out response to the voice messaging of user.It is corresponding with the personal identification characteristic information of different user by obtainingVirtual Intelligent customer service image carries out response to the voice messaging of user, can promote the personalized shape of Virtual Intelligent customer service imageAs enhancing the individualized experience of user.
Optionally, the answer gender and answer audio frequency characteristics, answer affective characteristics and Virtual Intelligent customer service that will acquireImage characteristics carry out integrated can be understood as the composition of task image characteristics corresponding to modules to total personage's shapeAs in so that the Virtual Intelligent customer service image integrated it is more life-like with naturally, and also original subscriber interacted with machine it is trueProperty.
Step S290: the Virtual Intelligent customer service is shown.
The matched method of intelligent customer service provided in this embodiment is realized the voice messaging for obtaining user the case whereUnder, the personal identification characteristic information of user is identified according to voice messaging, and then can pass through according to personal identification characteristic informationDifferent upgrade of network models identifies corresponding answer feature, these answer features are being integrated into a Virtual Intelligent visitorClothes, so that realizing that the corresponding Virtual Intelligent customer service of matching is shown carries out response with the voice messaging to user, so that userIt is more true to nature with the interactive mode of robot, the vividness and interest of human-computer interaction are promoted, and then enhance user experience.
As shown in figure 5, the method flow diagram for the matched method of intelligent customer service that the another embodiment of the application provides is shown,The described method includes:
Step S310: the voice messaging of user is received.
It is understood that in public place terminal device for, when user by these terminal devices intoWhen row human-computer interaction, inevitably it is mingled with noise, including ambient sound in the voice messaging of accessed user, reverberation,For example, the voice data etc. of other users.
So, in this case, terminal device can first receive the voice messaging of user.Optionally, the voice messagingMay include the voice data for attempting to carry out the user of human-computer interaction with terminal device, and the user that passes by or with attempt withTerminal device carries out the voice data of the companion of the user of human-computer interaction.
Step S320: identify whether the voice messaging is that reference sound source user issues.
As a kind of mode, in order to realize accurate human-computer interaction, the voice messaging of received user can be sievedWhether choosing is that reference sound source user sending identifies to these voice messagings, wherein reference sound source user can be machinePeople's customer service is initially received that user of voice messaging.Optionally, that position that robot customer service will can make a sound at firstUser is identified as reference sound source user, and then records to the sound of the user, and subsequent is known the sound of the userNot, the voice data of automatic fitration other users.
Optionally, in order to avoid misrecognition, that maximum user of sound of speaking can also be identified as by robot customer serviceReference sound source user, it is to be understood that the sound of speaking that the user of human-computer interaction is generally directly carried out with terminal device is bigUser's other users passed by other words in side, and the sound for the user that terminal device is spoken is faced than side pairOr the sound of user spoken far from terminal device it is big.
Optionally, the present embodiment can be unrestricted for the identification method of reference sound source user, can be according to practical feelingsCondition is set.
Step S330: if so, using the voice messaging as the voice messaging in interactive process.
As a kind of mode, if recognizing certain voice messaging for reference sound source user sending, then the voice can be believedBreath is as the voice messaging in interactive process.
For example, in a specific application scenarios, it is assumed that there is a user to come together with her several good friends a certainTerminal device information inquiry made above with human-computer interaction function, the user have said sentence " coming here " at first, thatThe user will be identified as reference sound source user.If then her good friend also starts, " what's the matter, comes here what hasDoes is thing, this enquiry machine route can be looked into or not ", at this moment, which suspends and does not talk to listen the speech of friend,After the friend of the user finishes speaking, which loquiturs again, then terminal device will be interior by speaking for the userHold as the voice messaging in interactive process.
Step S340: the voice messaging is inputted into the first nerves network model that training obtains in advance, identifies correspondenceUser personal identification characteristic information.
Step S350: corresponding Virtual Intelligent customer service is matched to the voice messaging according to the personal identification characteristic informationCarry out response.
Step S360: the Virtual Intelligent customer service is shown.
Step S370: if it is not, then abandoning the voice messaging.
It can be understood that in order to reduce the complexity of calculating, and storage redundancy is reduced, it can will not be above-mentioned useThe voice messaging at family is abandoned, so as to accelerate arithmetic speed.
The matched method of intelligent customer service provided in this embodiment is realized the voice messaging for obtaining user the case whereUnder, the voice messaging of user is identified, when the voice messaging for recognizing user is with reference to the case where the voice messaging of userUnder, using the voice messaging as the voice messaging in interactive process, the personal identification of user is then identified according to voice messagingCharacteristic information, and then corresponding Virtual Intelligent customer service can be matched according to personal identification characteristic information and shown to user'sVoice messaging carries out response, so that the interactive mode of user and robot is more true to nature, promotes the vividness and interest of human-computer interactionTaste, and then enhance user experience.
As shown in fig. 6, the method flow diagram of the matched method of intelligent customer service of the application another embodiment offer is provided,The described method includes:
Step S410: the voice messaging of user in interactive process is obtained.
Step S420: the voice messaging is inputted into the first nerves network model that training obtains in advance, identifies correspondenceUser personal identification characteristic information.
Step S430: judge the personal identification characteristic information whether be target user personal identification characteristic information.
Wherein, target user, which can be, registers with the user of account number in terminal device in advance.As a kind of mode, for oneA little personalized intelligent customer service services, for example, the backstage of public platform puts question to and explains the puzzle, in order to promote the letter in human-computer interaction processThe safety and privacy of breath, could normal use in the case where needing user's registration account number and logging in account number.So, thisIn the case of, it, can be with after the voice messaging in the interactive process according to acquisition identifies the personal identification characteristic information of userIt whether is that (wherein, the personal information of target user can for the personal identification characteristic information of target user to personal identification characteristic informationWith associated storage corresponding with account when target user registers account number) judge, in order to detect the individual of active userWhether identity characteristic information meets secret interaction permission.
As a kind of mode, personal identification characteristic information addition corresponding for account number can be used in user's registration account numberFamily mark, the mark can uniquely distinguish personally identifiable information corresponding with a certain account number with user.When obtaining active user'sAfter personal identification characteristic information, it can first judge the personal identification characteristic information with the presence or absence of above-mentioned mark, and if it exists, optional, can judge again the mark it is corresponding with target user's account number identify whether it is consistent, and then can be according to judging result to obtainingThe personal identification characteristic information of the active user taken is handled.
Step S440: if so, matching corresponding Virtual Intelligent customer service to institute's predicate according to the personal identification characteristic informationMessage breath carries out response.
It is understood that if the personal identification characteristic information of active user is the personal identification feature letter of target userBreath, answers the voice messaging of user then can match corresponding Virtual Intelligent customer service according to the personal identification characteristic informationIt answers, wherein concrete implementation process is referred to the description in previous embodiment, and details are not described herein.
Step S450: the Virtual Intelligent customer service is shown.
Step S460: if it is not, then abandoning the personal identification characteristic information.
Optionally, if the personal identification characteristic information of active user is not the personal identification characteristic information of target user, thatThe personal identification characteristic information can be then abandoned, so as to reduce the data storage capacity of terminal device, while promoting identificationAccuracy.
The matched method of intelligent customer service provided in this embodiment is realized the voice messaging for obtaining user the case whereUnder, the personal identification characteristic information of user is identified according to voice messaging, then carry out to the personal identification characteristic information identifiedVerifying, that is, judge personal identification characteristic information whether be target user personal identification characteristic information, if so, according to a personPart characteristic information, which matches corresponding Virtual Intelligent customer service and shown, carries out response with the voice messaging to user so that user withThe interactive mode of robot is more true to nature, promotes the vividness and interest of human-computer interaction, and then enhance user experience.
As shown in fig. 7, the method flow diagram of the matched method of intelligent customer service of the application another embodiment offer is provided,The described method includes:
Step S510: demand inquiry message is issued the user with by preset standard intelligent customer service model.
Optionally, for Virtual Intelligent customer service, if will until user speak just go to the voice messaging of user intoRow response can seem less friendly.On the one hand, it for from the user that remarkable machine interactive function is not used, may be unaware thatHow this, which goes, interacts, such as, it is not known that it directly inputs text or number or presses what button to carry out subsequent behaviourMake etc.；On the other hand, if robot customer service at the beginning is not spoken, after user speaks, robot customer service is abruptly started toIt speaks, user may be caused to frighten, then, the machine for the friendly performance of hoisting machine people's customer service, in the present embodimentPeople's customer service can actively issue the user with demand inquiry message before user speaks.
Wherein, standard can only customer service model can be understood as general, robot that is more standardizing customer service image.ExampleSuch as, the virtual customer service image of the neat and artistic generous women of tooling, instrument is worn, in other words the capable and experienced male of posture standard, spiritVirtual customer service image.Optionally, the specific setting means of standard intelligent customer service model can be changed according to the actual situation, exampleSuch as, the standard intelligent customer service image of different industries can be different.So, as a kind of mode, robot customer service can pass throughPreset standard intelligent customer service model issues the user with demand inquiry message, for example, actively inquiring whether the user needs to assortedHelp etc., it is diversified specifically to inquire that content can be, does not illustrate and is not construed as limiting one by one herein.
Step S520: the voice messaging of user in interactive process is obtained.
Step S530: the voice messaging is inputted into the first nerves network model that training obtains in advance, identifies correspondenceUser personal identification characteristic information.
Step S540: corresponding Virtual Intelligent customer service is matched to the voice messaging according to the personal identification characteristic informationCarry out response.
Step S550: the Virtual Intelligent customer service is shown.
The matched method of intelligent customer service provided in this embodiment, by preset standard intelligent customer service model actively toUser issues demand inquiry message, then obtains the voice messaging of user in interactive process, then inputs voice messaging preparatoryThe first nerves network model that training obtains, identifies the personal identification characteristic information of corresponding user, further according to personal identificationCharacteristic information matches corresponding Virtual Intelligent customer service and carries out response to voice messaging, then shows Virtual Intelligent customer service.By upperThe mode of stating is realized in the case where obtaining the voice messaging of user, identifies that the personal identification of user is special according to voice messagingReference breath, and then corresponding Virtual Intelligent customer service can be matched according to personal identification characteristic information and shown with the language to userMessage breath carries out response, so that the interactive mode of user and robot is more true to nature, promotes the vividness and entertaining of human-computer interactionProperty, and then enhance user experience.
As shown in figure 8, showing the structural block diagram of the matched device 600 of intelligent customer service provided by the embodiments of the present application, instituteState device 600 and run on the terminal device with display screen or other image output devices, terminal device can be smart phone,The electronic equipments such as tablet computer, wearable intelligent terminal, described device 600 include:
Module 610 is obtained, for obtaining the voice messaging of user in interactive process.
Wherein, the acquisition module 610 may include receiving unit, which is used to receive the voice letter of userBreath；Recognition unit, whether the voice messaging is that reference sound source user issues to the recognition unit for identification；Voice messaging obtainsUnit, the voice messaging acquiring unit are used for if so, using the voice messaging as the voice messaging in interactive process.
As a kind of mode, device 600 can also include that sample set obtains module, which obtains module for obtainingTraining sample set, the training sample concentrate include the training audio extracted from training voice data and with the trained soundFrequently corresponding personal identification characteristic information, the personal identification characteristic information include at least the gender of user, age, accent, stateNationality, emotion information；Network model obtains module, which obtains module and be used to the training sample set being input to machineLearning model, and the machine learning model is trained, obtain first nerves network model.
As a kind of mode, described device 600 can also include judgment module, and the judgment module is for judging the individualWhether identity characteristic information is the personal identification characteristic information of target user, if so, executing described according to the personal identificationCharacteristic information matches corresponding Virtual Intelligent customer service and carries out response to the voice messaging, if it is not, then abandoning the described personPart characteristic information.
As a kind of mode, described device 600 can also include inquiry module, and the inquiry module is for obtaining interactive processBefore the voice messaging of middle user, demand inquiry message is issued the user with by preset standard intelligent customer service model.
Identification module 620, for the voice messaging to be inputted the first nerves network model that training obtains in advance, identificationThe personal identification characteristic information of corresponding user out.
Matching module 630, for matching corresponding Virtual Intelligent customer service to described according to the personal identification characteristic informationVoice messaging carries out response.
As a kind of mode, matching module 630 may include first acquisition unit, special with the personal identification for obtainingReference ceases corresponding answer gender and replies audio frequency characteristics, and the answer audio frequency characteristics, which include at least, replies used languageSound, intonation, language；Second acquisition unit, for obtaining answer affective characteristics corresponding with the personal identification characteristic information；TheThree acquiring units, for obtaining Virtual Intelligent customer service image characteristics corresponding with the personal identification characteristic information；Response unit,For by preset algorithm by the answer gender and reply audio frequency characteristics, the answers affective characteristics and it is described virtuallyIntelligent customer service image characteristics are integrated, and are generated Virtual Intelligent customer service and carried out response to the voice messaging.
Wherein, as an implementation, the first acquisition unit is specifically used for the personal identification characteristic informationIt is input to the nervus opticus network model that training obtains in advance, obtain corresponding answer gender and replies audio frequency characteristics.
In one implementation, the first acquisition unit may include detection sub-unit, which is used forThe voice messaging is detected with the presence or absence of voice variation characteristic；Feature obtains subelement, if this feature obtains subelement for depositingThe personal identification characteristic information is being input to the nervus opticus network model that training obtains in advance, is being obtained and the voiceThe answer gender and answer audio frequency characteristics of variation characteristic adaptation.
As an implementation, the second acquisition unit is specifically used for for the personal identification characteristic information being input toThe third nerve network model that training obtains in advance, obtains corresponding answer affective characteristics, wherein the third nerve network mouldType is different from the nervus opticus network model.
Display module 640, for showing the Virtual Intelligent customer service.
The matched device of intelligent customer service provided in this embodiment, by obtaining the voice messaging of user in interactive process, afterAnd voice messaging is inputted into the first nerves network model that training obtains in advance, identify the personal identification feature of corresponding userInformation matches corresponding Virtual Intelligent customer service further according to personal identification characteristic information and carries out response to voice messaging, then showsVirtual Intelligent customer service.It is realized through the above way in the case where obtaining the voice messaging of user, is known according to voice messagingNot Chu user personal identification characteristic information, and then corresponding Virtual Intelligent customer service can be matched according to personal identification characteristic informationIt is shown and response is carried out with the voice messaging to user, so that the interactive mode of user and robot is more true to nature, promote peopleThe vividness and interest of machine interaction, and then enhance user experience.
The matched device of intelligent customer service provided by the embodiments of the present application is for realizing corresponding intelligence in preceding method embodimentThe energy matched method of customer service, and the beneficial effect with corresponding embodiment of the method, details are not described herein.
It is apparent to those skilled in the art that video process apparatus provided by the embodiments of the present application canRealize each process in preceding method embodiment, for convenience and simplicity of description, the specific work of foregoing description device and moduleMake process, can be refering to the corresponding process in preceding method embodiment, details are not described herein.
In embodiment provided herein, the mutual coupling of shown or discussed module or direct-coupling orCommunication connection can be through some interfaces, and the indirect coupling or communication connection of device or module can be electrical property, it is mechanical or itsIts form.
In addition, each functional module in the embodiment of the present application can integrate in a processing module, it is also possible to eachA module physically exists alone, and can also be integrated in a module with two or more modules.Above-mentioned integrated module was bothIt can take the form of hardware realization, can also be realized in the form of software function module.
Referring to Fig. 9, it illustrates a kind of structural block diagrams of terminal device 101 provided by the embodiments of the present application.The terminalEquipment 101, which can be smart phone, tablet computer, e-book etc., can run the terminal device of application program.In the applicationTerminal device 101 may include one or more such as lower component: processor 202, memory 204 and one or more application journeySequence, wherein one or more application programs can be stored in memory 204 and be configured as by one or more processors202 execute, and one or more programs are configured to carry out the method as described in preceding method embodiment.
Processor 202 may include one or more processing core.Processor 202 is whole using various interfaces and connectionVarious pieces in a terminal device 101, by run or execute the instruction being stored in memory 204, program, code set orInstruction set, and the data being stored in memory 204 are called, execute the various functions and processing data of terminal device 101.It canSelection of land, processor 202 can use Digital Signal Processing (Digital Signal Processing, DSP), field-programmableGate array (Field-Programmable Gate Array, FPGA), programmable logic array (Programmable LogicArray, PLA) at least one of example, in hardware realize.Processor 202 can integrating central processor (CentralProcessing Unit, CPU), in image processor (Graphics Processing Unit, GPU) and modem etc.One or more of combinations.Wherein, the main processing operation system of CPU, user interface and application program etc.；GPU is for being responsible forShow the rendering and drafting of content；Modem is for handling wireless communication.It is understood that above-mentioned modemIt can not be integrated into processor 202, be realized separately through one piece of communication chip.
Memory 204 may include random access memory (Random Access Memory, RAM), also may include read-onlyMemory (Read-Only Memory).Memory 204 can be used for store instruction, program, code, code set or instruction set.It depositsReservoir 204 may include storing program area and storage data area, wherein the finger that storing program area can store for realizing operating systemEnable, for realizing at least one function instruction (such as touch function, sound-playing function, image player function etc.), be used forRealize the instruction etc. of following each embodiments of the method.Storage data area can be created in use with storage terminal device 101Data (such as phone directory, audio, video data, chat record data) etc..
Referring to Fig. 10, it illustrates a kind of structural frames of computer readable storage medium provided by the embodiments of the present applicationFigure.Program code is stored in the computer readable storage medium 700, said program code can call execution above-mentioned by processorMethod described in embodiment of the method.
Computer readable storage medium 700 can be such as flash memory, EEPROM (electrically erasable programmable read-only memory),The electronic memory of EPROM, hard disk or ROM etc.Optionally, computer readable storage medium 700 includes non-volatile meterCalculation machine readable medium (non-transitory computer-readable storage medium).Computer-readable storageMedium 700 has the memory space for the program code 710 for executing any method and step in the above method.These program codes canWith from reading or be written in one or more computer program product in this one or more computer program product.Program code 710 can for example be compressed in a suitable form.
In conclusion the matched method, apparatus of intelligent customer service provided by the embodiments of the present application, terminal device and storage are situated betweenVoice messaging is then inputted the first nerves that training obtains in advance by obtaining the voice messaging of user in interactive process by matterNetwork model identifies the personal identification characteristic information of corresponding user, matches further according to personal identification characteristic information correspondingVirtual Intelligent customer service carries out response to voice messaging, then shows Virtual Intelligent customer service.It realizes and is obtaining through the above wayIn the case where the voice messaging of user, the personal identification characteristic information of user is identified according to voice messaging, and then can rootCorresponding Virtual Intelligent customer service is matched according to personal identification characteristic information shown and response is carried out with the voice messaging to user, makeThe interactive mode for obtaining user and robot is more true to nature, promotes the vividness and interest of human-computer interaction, and then enhance user's bodyIt tests.
Finally, it should be noted that above embodiments are only to illustrate the technical solution of the application, rather than its limitations；AlthoughThe application is described in detail with reference to the foregoing embodiments, those skilled in the art are when understanding: it still can be withIt modifies the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；AndThese are modified or replaceed, do not drive corresponding technical solution essence be detached from each embodiment technical solution of the application spirit andRange.