Specific embodiment
Technical solution for a better understanding of the present invention with reference to the accompanying drawing retouches embodiments herein in detail
It states.
It will be appreciated that described embodiment is only a part of the embodiment of the application, instead of all the embodiments
Enumerate.Based on embodiment described in the disclosure, those of ordinary skill in the art create the feelings of habit labour not paying
Every other change case obtained belongs to the protection scope of the application under condition.
In existing customer service system, fixed music or busy is often played in the situation for staying in phone such as busy repeatedly
Sound, or the often simple waiting in the situation of chat.System does not make full use of this period to obtain useful letter
It ceases and/or takes strategy to improve user satisfaction.
For example, Fig. 1 shows the flow chart of the customer service worksheet processing method 100 of the prior art.In frame 102, from user receive to
The request that customer service system is initiated.For example, user can dial service calls, or initiate the chat etc. with online customer service.In frame
104, customer service robot guides user's exposition need.For example, customer service robot can be putd question to directly to user in the situation of phone
To obtain the answer of user's Yes/No, or client can be guided to select to go for taking in services menu by key etc.
The problem of business major class.And in the situation of chat, customer service robot can be putd question to directly to user to obtain returning for user's Yes/No
It answers, or user can be guided directly to describe problem, or options for user is provided and clicks the ready-made problem major class etc. of selection.Example
Such as, in an exemplary and non-limiting shopping online scene, customer service robot can inquire that user is intended to it is to be understood that commodity are believed
It ceases, requires goods return and replacement, is still other.In another exemplary and non-limiting handling bank business scene, customer service robot
Can inquire that user is individual client or corporate client, be intended to open an account, handle the deposit or the withdrawal, purchase finance product or its
He.The above is only the example of customer service robot guidance user's exposition need, the disclosure is not limited fixed in this regard.
Figure 1A shows the example scenario during the customer service worksheet processing of the prior art.(A) of Figure 1A shows such as net
Shop shopping scene, wherein customer service robot inquiry user wants to know about merchandise news or goods return and replacement etc., and then user proposes
The hope of goods return and replacement.(B) of Figure 1A, which is shown, for example handles telephone banking scene, and wherein customer service robot inquiry user thinks
It opens an account, handle the deposit or the withdrawal, purchase finance product or other, then user indicates oneself by way of key
Demand.(C) of Figure 1A shows such as APP scene, and wherein customer service robot guidance user expresses the demand of oneself, then user
Inform the problem of oneself cannot log in original account because changing cell-phone number.
According to roughly determining customer problem major class, in frame 106, it is further right that customer service robot is carried out with user
Words, for example to refine user demand and/or obtain further useful information.For example, in the situation of phone, customer service robot
It can provide common several problems to select for user.
Figure 1B shows the example scenario during the customer service worksheet processing of the prior art.For example, (A) with Figure 1A is corresponding
Ground, (A) of Figure 1B are shown for example in aforementioned shopping online scene, when determining that user has selected goods return and replacement to service, can be ask
Ask whether user is service below needing, such as: (i) is only intended to understand goods return and replacement policy;(ii) it needs to return goods;(iii) it needs to change
Goods;Or (iv) is other, then user expresses the intention to be returned goods.For another example, accordingly with (B) of Figure 1A, (B) of Figure 1B shows
Go out and handled in telephone banking scene aforementioned, when determining that user has selected individual to open an account, can ask the user whether to need
It to service below, such as (i) open up RMB current account;(ii) foreign currency current account is opened up;(iii) it is regular to open up RMB
Account;(iv) foreign currency fixed account is opened up;Or it is (v) other.In such scene, bootable client continues through key etc. and exists
Required specific service content etc. is further selected in next stage services menu.For example, in this instance, user's selection opens up the people
Coin current account.In the non-limiting scene of another exemplary, for example, accordingly with (C) of Figure 1A, (C) of Figure 1B shows
Gone out customer service robot can continue guide user problem is described, such as " you are that can not log in original account now? " and
User may directly make voice or typewriting is replied.Customer service robot can for example, by crawl user reply in keyword,
Such as cell-phone number, login, loss etc. are come details the problem of further determining that user.The above is only customer service robots and user to carry out
The example further talked with, the disclosure are not limited fixed in this regard.
In frame 108, system determines whether that user demand or customer service can be analyzed according to collected user information
Whether robot and user session have exceeded limitation.Determining whether can be based on scheduled time threshold, Huo Zheke beyond limitation
With the wheel number (for example, at most three-wheel) etc. based on customer service robot and user session, or any combination thereof.If according to frame
The information obtained in 104 and 106 has been able to analyze user demand or customer service robot is alreadyd exceed with user session
Such as threshold time but still not can determine that the specific requirements of user, then method 100 goes to frame 110.For example, Figure 1A (A) and
In the scene of (B) of (A) of Figure 1B and (B) of Figure 1A and Figure 1B, system may be had been able to from the reply that user provides
Accurately determine user demand.Then method 100 may be advanced to frame 110.
On the other hand, if not can determine that user is intended to and customer service robot and user session not yet exceed limitation still,
Such as in aforementioned on-line shop's shopping scene, user has selected (iv) other, or in aforementioned handling bank business scene, Yong Huxuan
Select that (v) is other, this shows intelligent robot not yet and can be recognized accurately user's intention, then method 100 returns to frame 106, herein
Customer service robot and user carry out further dialogue to attempt to determine that user is intended to.For another example, in (C) of Figure 1A and Figure 1B
(C) in APP scene, user still without sufficiently clear gives expression to pair of his/her intention but customer service robot and user
Words are not yet beyond limitation, then method 100 returns to frame 106.
In frame 110, it is intended to based on identified user and user is assigned to properly by other associated information, system
Manual service technical ability group further to be talked with by artificial customer service with user and User Demand.For example, when in frame 108
It determines in the case where having analyzed user demand, can be assigned to user suitably based on the user demand analyzed
Manual service technical ability group.For example, user can be assigned to by system to be responsible for moving back in the scene of (A) of (A) and Figure 1B of Figure 1A
The artificial customer service technical ability group of goods.For another example, in the scene of (B) of Figure 1A and (B) of Figure 1B, user can be assigned to open by system
If the artificial customer service technical ability group of RMB current account.According to another example, when in frame 108 because of customer service robot and user couple
Words have exceeded limitation (for example, customer service robot and user session have been more than scheduled time threshold or customer service robot and have used
The wheel number of family dialogue has been more than scheduled wheel number threshold value etc., or any combination thereof) and in the case where entering frame 110, system
User can be assigned to and be responsible for difficult miscellaneous artificial customer service technical ability group.
In frame 114, service is provided for user by the artificial customer service in the suitable manual service technical ability group assigned.So
And the customer service of nobody work is idle in manual service technical ability group if appropriate, then between frame 110 and 114, it is also possible to deposit
In busy waiting frame 112.In busy waiting frame 112, user can only passively wait the service of artificial customer service to be obtained.System is extremely
Mostly user plays music or busy tone, without making full use of obtain useful information this period, or takes tactful increase
User satisfaction.Even in the system having, in the case where the busy waiting time is overtime, system can also terminate automatically and user
Session.For example, being more than scheduled time threshold value when the busy waiting time in the scene of phone customer service, system will be automatically hung up
Phone.It is excessively poor experience for a user by on-hook in the case where having answered many problems and having waited for a good while again.?
In another exemplary situation, user is assigned to the artificial of inaccuracy since customer service robot has exceeded limitation with user session
Customer service technical ability group often finds that user should be assigned in waiting and after the artificial customer service inquiry in artificial customer service technical ability group
Another artificial customer service technical ability group.To again user needs the waiting into next round.This is inefficiency and user experience is bad
's.
Fig. 2 shows the flow charts of customer service intelligence worksheet processing method 200 according to the one side of the disclosure.With 100 class of method
As, in frame 202, the request initiated to customer service system is received from user.With the frame 102 of method 100 the difference is that, frame
202 can also guess user demand according to the characterization factor (for example, action trail etc. of user) of user.For example, user can dial
Service calls are beaten, or initiate the chat etc. with online customer service.According to an exemplary and non-limiting embodiment, for example, user
Action trail show that user's other day has purchased some article, and according to logistics information, which is not yet sent to.In this feelings
In shape, system can guess that user may want to inquire when the article can be sent to.For another example, the action trail of user shows user
Other day has purchased some article, and according to logistics information, which has been sent to.In this case, system can guess that user can
It can want goods return and replacement.For another example, the action trail of user shows user and repeatedly browses certain class article within the next few days but not yet place an order.?
In this situation, system can guess that user may wish to inquire details of the article etc..It is non-limiting according to another exemplary
Embodiment, frame 202 can be omitted.
In frame 204, customer service robot guides user's exposition need.For example, customer service robot can be straight in the situation of phone
It connects to user and puts question to obtain the answer of user's Yes/No, or user can be guided to select in services menu by key etc.
The problem of going for service major class.And in the situation of chat, customer service robot can be putd question to directly to user to obtain use
The answer of family Yes/No, or user can be guided directly to describe problem, or provide options for user and click the ready-made problem of selection
Major class etc..For example, customer service robot can inquire that user is intended in an exemplary and non-limiting shopping online scene
Solution merchandise news requires goods return and replacement, is still other.In another exemplary and non-limiting handling bank business scene, visitor
Taking robot can inquire that user is individual client or corporate client, be intended to open an account, handle the deposit or the withdrawal, purchase financing production
Product or other.In the case where frame 202 guesses user demand, also user can be guided based on the user demand guessed
Exposition need.For example, being possible in the situation for wanting to inquire certain special article guessing user, it can preferentially inquire that client is
The no favor information etc. for wanting to know about the particular commodity and/or the particular commodity.For another example, it is possible to want to move back guessing user
In the situation exchanged goods, it can preferentially ask the user whether to have the cargo dissatisfied etc..The above is only customer service robots to draw
The example of user's exposition need is led, the disclosure is not limited fixed in this regard.
According to roughly determining customer problem major class, in frame 206, it is further right that customer service robot is carried out with user
Words, for example to refine user demand and/or obtain further useful information.For example, in the situation of phone, customer service robot
It can provide common several problems to select for user.For example, in aforementioned shopping online scene, when determining that user selected to replace
When goods services, it can ask the user whether to be service below needing, such as: (i) is only intended to understand goods return and replacement policy;(ii) it needs to move back
Goods;(iii) it needs to exchange goods;Or (iv) is other.For another example, in aforementioned handling bank business scene, when determining that user selects
When individual opens an account, following service can be inquired whether the user needs to, such as (i) opens up RMB current account;(ii) foreign currency is opened up
Current account;(iii) RMB fixed account is opened up;(iv) foreign currency fixed account is opened up;Or it is (v) other.In such scene
In, bootable client continues through key etc. and further selects required specific service content etc. in next stage services menu.
In the non-limiting scene of another exemplary, customer service robot can directly guide user to describe problem, such as " you are good,
What, which be may I ask, can help you? " and user can directly voice or typewriting reply: " cell-phone number is lost, and is then mended again
Do, then log in again APP just with it is pervious what if be not an account? ".Customer service robot can be used for example, by crawl
Keyword, such as account, login, loss in the reply of family etc. are come details the problem of determining user.The above is only customer service robots
The example further talked with user, the disclosure are not limited fixed in this regard.It is non-limiting according to another exemplary
Embodiment, frame 204 and 206 can also be merged, as long as being able to carry out the dialogue of customer service robot and user to guide user's table
Up to demand or provide relevant information.
In frame 208, system determines whether that user demand or customer service can be analyzed according to collected user information
Whether robot and user session have exceeded limitation.Determining whether can be based on scheduled time threshold, Huo Zheke beyond limitation
With the wheel number etc. based on customer service robot and user session, or any combination thereof.If according to what is obtained in frame 204 and 206
Information, have been able to analyze user demand or customer service robot and user session alreadys exceed such as threshold time but still
The specific requirements of user are not can determine that, then method 200 goes to frame 210.If cannot still analyze user demand and service machine
Device people and user session not yet exceed limitation, then method 200 returns to frame 206, carry out in this customer service robot and user further
Dialogue with attempt determine user be intended to.It is clicked in the situation for selecting ready-made problem major class for example providing options for user,
If user continuously selects " other " option, this, which shows intelligent robot not yet, can be recognized accurately user's intention.For another example, in visitor
It takes in the situation that robot can be putd question to directly to user to obtain the answer of user's Yes/No, user continuously provides negative acknowledge, then
User's intention can be recognized accurately by showing intelligent robot not yet.For another example, in the situation that guidance user directly describes problem, such as
Fruit extracts effective keyword in cannot describing from user, then shows that intelligent robot fails that user's intention is recognized accurately.
In frame 210, it is intended to based on identified user and user is assigned to by other associated information, system automatically
Suitable manual service technical ability group with user further talk with simultaneously User Demand by artificial customer service.For example, working as
In the case that the determination of frame 108 has analyzed user demand, user can be assigned to based on the user demand analyzed
Suitable manual service technical ability group.According to another example, when in frame 108 because customer service robot and user session have exceeded limitation
(for example, customer service robot and user session have been more than the wheel number of scheduled time threshold or customer service robot and user session
It has been more than scheduled wheel number threshold value etc., or any combination thereof) and in the case where entering frame 110, system can refer to user
It sends to and is responsible for difficult miscellaneous artificial customer service technical ability group.
In frame 214, service is provided for user by the artificial customer service in the suitable manual service technical ability group assigned.Method
200 be in place of another difference of method 100, the customer service of nobody work is empty in manual service technical ability group if appropriate
Spare time, then the busy waiting frame between frame 210 and 214 is personalized busy waiting frame 212.In personalized busy waiting frame 212
In, system, which can make full use of, to be continued to obtain useful information this period, or strategy is taken to increase user satisfaction.According to one
Exemplary embodiment, the customer service of nobody work is idle in manual service technical ability group if appropriate, and system prediction user
The service of artificial customer service cannot be obtained within threshold time (for example, 2 minutes), then system enters personalized busy waiting frame
212;Otherwise, if prediction user can obtain the service of artificial customer service within threshold time, directly user is allowed to continue waiting for
Until obtaining artificial customer service.The nonlimiting examples according to another exemplary, system can also be directly entered personalized numerous
Busy waiting frame 212.The personalized busy specific embodiment for waiting frame 212 is described below in association with Fig. 2A and 3 etc..
Fig. 2A is shown according to the disclosure on the one hand to the exemplary realization 200A of the busy waiting frame 212 of the personalization of Fig. 2.
And nonlimiting examples exemplary according to one, method 200A may include in frame 202A, can for user and/or customer problem/
Demand carries out portrait modeling (profiling).For example, to user carry out portrait modeling may include to user carry out age bracket,
The classification of gender, occupation, residence, educational background, business history etc..For another example, portrait modeling can be carried out to user demand.Such as to
It may include carrying out the classification of urgency level or business demand that family demand, which carries out portrait modeling,.For example, urgency level classification can be divided into
It is high, medium and low or be divided into finer 0-10 grade etc..The disclosure is not limited fixed in this regard.The classification of business demand can be by tool
Body business is divided into such as pre-sales, mid-sales, after sale.The disclosure is also not limited in this regard.
After frame 202A carries out portrait modeling for user and/or customer problem/demand, method 200A proceeds to frame
204A.In frame 204A, system determines whether the confidence level of user demand is sufficiently high.For example, determining the confidence level of user demand is
It is no it is sufficiently high may include determining user to the descriptions of problems/requirements whether sufficiently clear and/or information whether enough, really
Whether the artificial customer service technical ability group classification of the worksheet processing model output before determining accurately waits enough and/or any combination thereof.Determination is set
Reliability can be based at least partially on the portrait modeling carried out for user and/or customer problem/demand.If confidence level is enough
It is high, then it is assumed that information needed for description sufficiently clear and/or the accurate manual service of progress of the user to problems/requirements is
Through enough.Otherwise it is assumed that user is not clear enough to the description of problems/requirements and/or information is not enough.It is exemplary according to one and
Nonlimiting examples, for example, when in the frame 208 of Fig. 2 be because customer service robot and user session have exceeded limitation and in
In the situation that only customer service robot further talks with and assigned user with user, it may be determined that artificial customer service technical ability group classification
Confidence level it is not high enough.For example, in this case, it is believed that user is not clear enough to the descriptions of problems/requirements and/or information not
It is enough abundant.
If system judges that user is sufficient to the description sufficiently clear of problems/requirements and/or information in frame 204A
When enough, then method 200A may be advanced to frame 206A.According to an alternative, frame 202A can also be in the "Yes" branch of frame 204A
It is executed before frame 206A later.In the case, determine that confidence level can be not based on for user and/or customer problem/need
The portrait of progress is asked to model.For example, in the case, determine that confidence level can be based at least partially on is in the frame 208 of Fig. 2
It is no to go to frame 210 because customer service robot and user session are beyond limitation and assigned automatically.
In frame 206A, system can recommend the waiting pastime for being adapted to user in busy waiting time based on the portrait of user.
For example, system can provide intelligent sound box service.Intelligent sound box service for example may include playing recommend music, provide music for user
Program request provides and chats the service such as robot.Music recommends to draw a portrait based on user and the Music on Demand of the user is believed in history
Breath etc., is realized using above-described proposed algorithm and model etc..When there is the artificial customer service free time, the waiting of end block 206A
Pastime (for example, chat of music and/or chat robot), then goes to the frame 214 of the method 200 of Fig. 2.
However, when in frame 204A, when system judges that user is not clear enough to the description of problems/requirements and/or information is inadequate,
Method 200A may be advanced to frame 208A.Needs further inquiry user can be determined in frame 208A, method 200A.Show according to one
Example property and nonlimiting examples, when stopping customer service robot and user due to customer service robot and user session have exceeded limitation
Further dialogue and user is assigned and is determined in not high enough etc. the situation of the confidence level of artificial customer service technical ability group classification, it can
Think that user is not clear enough to the description of problems/requirements and/or information is not enough.In the case, for example, although will use
Family is assigned to specific artificial customer service technical ability group (for example, being responsible for difficult miscellaneous artificial customer service technical ability group or another inaccuracy
Artificial customer service technical ability group), but if there is the more time to user proposes that further problem and/or request user are provided into one
The information of step then probably enables a customer to be assigned to more suitable artificial customer service technical ability group, to obtain more acurrate
Service and/or shorten needed for the manual service time.Therefore, method 200A then returnes to the frame 206 of the method 200 of Fig. 2, from
And customer service robot is further talked with user, for example to obtain further useful information.To which system can be into
And turn again to frame 208, with determined whether to analyze according to the user information being further collected into user demand or
Whether customer service robot and user session have exceeded the automaticmanual customer service worksheet processing for limiting and correspondingly carrying out again.According to another
Alternative embodiment, can also further talk in customer service robot and user with obtain after further useful information without
Worksheet processing again, but subsequent manual service is carried out using useful information obtained.It, can according to another replacement new embodiment
Further talk in customer service robot and user and carries out selectively worksheet processing after further useful information to obtain.For example,
System can determine whether the accuracy of worksheet processing for the first time, and and if only if just worksheet processing again when worksheet processing departure degree is excessive for the first time.For example, if
User is assigned to close manual service technical ability group by worksheet processing for the first time, then can not carry out worksheet processing again;And if sent for the first time
Singly fail for user to be assigned to close manual service technical ability group, then can carry out worksheet processing again.
Further inquiry user is needed determining, the case where confidence level for example to improve the classification of artificial customer service technical ability group
Under, by proposing that further problem and/or request user provide further information to user using busy waiting period, just
It can be effectively by customers assignment to more suitable artificial customer service technical ability group, to more accurately be serviced and/or be shortened institute
Need the manual service time.
On the one hand Fig. 3 is shown to be realized the busy another exemplary for waiting frame 212 of the personalization of Fig. 2 according to the disclosure
300.And nonlimiting examples exemplary according to one, method 300 may include that can ask for user and/or user in frame 302
Topic/demand carries out portrait modeling (profiling).For example, carrying out portrait modeling to user may include carrying out the age to user
The classification of section, gender, occupation, residence, educational background, business history etc..For another example, portrait modeling can be carried out to user demand.Such as
Carrying out portrait modeling to user demand may include carrying out the classification of urgency level or business demand.For example, urgency level classification can
It is divided into high, medium and low or is divided into finer 0-10 grade etc..The disclosure is not limited fixed in this regard.The classification of business demand can
Be divided by specific business such as pre-sales, mid-sales, after sale.The disclosure is also not limited in this regard.
After frame 302 carries out portrait modeling for user and/or customer problem/demand, method 300 proceeds to frame
304.In frame 304, system determines whether the confidence level of user demand is sufficiently high.For example, determine user demand confidence level whether
It is sufficiently high to may include determining user whether enough whether sufficiently clear and/or information, determine to the descriptions of problems/requirements
Whether the artificial customer service technical ability group classification of worksheet processing model output before accurately waits enough and/or any combination thereof.Determine confidence
Degree can be based at least partially on the portrait modeling carried out for user and/or customer problem/demand.If confidence level is enough
It is high, then it is assumed that information needed for description sufficiently clear and/or the accurate manual service of progress of the user to problems/requirements is
Through enough.Otherwise it is assumed that user is not clear enough to the description of problems/requirements and/or information is not enough.It is exemplary according to one and
Nonlimiting examples, for example, when in the frame 208 of Fig. 2 be because customer service robot and user session have exceeded limitation and in
In the situation that only customer service robot further talks with and assigned user with user, it may be determined that artificial customer service technical ability group classification
Confidence level it is not high enough.For example, in this case, it is believed that user is not clear enough to the descriptions of problems/requirements and/or information not
It is enough abundant.
If system judges that user is sufficient to the description sufficiently clear of problems/requirements and/or information in frame 304
When enough, then method 300 may be advanced to frame 306.According to an alternative, frame 302 can also be after the "Yes" branch of frame 304
It is executed before frame 306.In the case, it determines that confidence level can be not based on to carry out for user and/or customer problem/demand
Portrait modeling.For example, in the case, determining that whether confidence level can be based at least partially in the frame 208 of Fig. 2 because of visitor
It takes robot and goes to frame 210 beyond limitation with user session and assigned automatically.
In frame 306, whether the problem of determining user is urgent.If urgent the problem of frame 306 determines user, enter frame
308.In frame 308, the portrait carried out for user and/or customer problem/demand can be based at least partially on and modeled to analyze use
The mood at family.It is exemplary according to one and nonlimiting examples, the mood for analyzing user can be realized by machine learning.Example
Such as, it can establish mood analysis model, which can be based on such as deep learning.The training of the mood analysis model can be
Have supervision, it is unsupervised or semi-supervised.And nonlimiting examples exemplary according to one, can be used supervised learning
To train the mood analysis model.This can be realized by providing the affection data collection through marking.It is exemplary according to one rather than
Limited embodiment, frame 308 can be incorporated into frame 302 or frame 302 can be incorporated into frame 308.Shown according to another
Example property and nonlimiting examples, frame 308 can also execute after frame 304 and before frame 306.The disclosure in this regard not by
It limits.
According to an exemplary embodiment, which can word based on such as user, punctuate, word speed, language
It adjusts etc. and the mood of user is analyzed.For example, there is more fierce word, exclamation mark, faster word speed, higher
Intonation etc. can be shown that user emotion excitement.According to a further exemplary embodiment, which can be for example based on the use
The mood of the user is analyzed in the history mood of the system at family.For example, if according to the user going through in the system
History mood, the general mood of the user is tranquil, has a friendly attitude, then can be shown that a possibility that user emotion is tranquil is higher.According to another
Exemplary embodiment, the mood analysis model can for example based on the user the system historical behavior come the mood to the user
It is analyzed.For example, if the user can be shown that user is emotional discontented once because of similar situation complaint or complaint etc..Feelings
Thread analysis model can also carry out mood analysis based on any combination of above-mentioned items.After trained, which for example can be with
Identify the various moods of user, such as whether eager user is, nervous, angry, irritated, discontented etc..
After analyzing the mood of user in frame 308, method 300 determines whether user emotion is negative in frame 310.Example
Such as, negative emotions may include but be not limited to angry, nervous, irritated, discontented, eager etc..When frame 310 determine user emotion be negative
Face, method 300 correspondingly improves the busy waiting priority of user in mood of the frame 312 based on user and portrait etc..According to
One is exemplary and nonlimiting examples, improve user busy waiting priority can for example including by user in waiting list
(for example, predetermined ranking or in advance predetermined percentage etc. in advance) in advance.The nonlimiting examples according to another exemplary,
It is higher that the busy waiting priority of raising user may also include experience level in the corresponding artificial customer service technical ability group of for example preferential appointment
Artificial customer service service the user.The nonlimiting examples according to another exemplary, the busy waiting for improving user are preferential
Grade may also include such as providing discount coupon or small gift to attempt to alleviate user's negative emotions.The disclosure is unrestricted in this regard
It is fixed.Then method 300 goes to frame 314.
When determining that user emotion is not negative (for example, being positive or neutral) in frame 310, method 300 goes to frame
314.In frame 314, the portrait based on user, system correspondingly recommends the waiting pastime for being adapted to user.It is exemplary according to one and
Nonlimiting examples, system can also recommend the waiting pastime for being adapted to user based on the mood of the user analyzed.Example
Such as, system can play the recommendation music being adapted with the mood of user automatically.The adaptable recommendation music with the mood of user
It can be determined based on the Music on Demand of the user in history, user's portrait, user's current emotional.For example, proposed algorithm can be used
Collaborative filtering.For example, the collaborative filtering based on user can find user to sound by the historical behavior data of user
The attitude (for example, preference) of happy (or other contents), and these attitudes can be measured and be given a mark.According to different user to phase
The attitude of same music (or other contents) calculates the relationship between user.Between the user for having identical attitude (for example, preference)
Carry out the recommendation of music (or other contents).For another example, the collaborative filtering based on content then can be by calculating different user pair
The scoring of different content (for example, music) is to obtain the relationship between content.User is carried out in similar based on the relationship between content
Hold the recommendation of (for example, music).Scoring can represent user to the attitude (for example, preference) of content.Based on collaborative filtering
Recommendation may include the collaborative filtering for example based on memory and the collaborative filtering based on model.Collaborative filtering based on model can be adopted
With matrix decomposition (Matrix Factorization) technology.For example, matrix decomposition technology can have m user, n content with
And sparse rating matrix R (R ∈ Rm*n) recommendation task in learn user-feature matrix U out and characteristic-article matrix V so that
R=UV (that is, the UV of matrix R is decomposed), so as to predict the scoring lacked in sparse rating matrix R.Collaboration based on model
Deep learning model, such as the Wide&Deep model of Google etc. can also be used in filtering, which passes through Wide model and Deep mould
The joint training of type can be remembered (that is, from the correlation found between interior perhaps feature in data is lasted) and extensive simultaneously
The ability of (i.e. the transmitting of correlation, to find that seldom in lasting data or even without appearance new feature combines).?
System played recommendation music automatically before/after/period, user can also voluntarily request music at any time.System continues to remember in turn
It records and the program request for accumulating the user is recorded with the training and update for recommended models.According to another exemplary embodiment, may be used
Chatting service etc. is provided based on the mood of the portrait of user and/or user by chat robots.When there is the artificial customer service free time,
The recommendation of end block 314 is diverted oneself, and the frame 214 of the method 200 of Fig. 2 is then gone to.
On the other hand, if it is not urgent the problem of frame 306 determines user, enter frame 312.In frame 312, system can be
Busy waiting time recommends the waiting pastime for being adapted to user.For example, system can provide intelligent sound box service.Intelligent sound box service
It such as may include playing recommend music, Music on Demand is provided, provides and chats the service such as robot for user.Music recommendation can be based on
User's portrait and in history the Music on Demand information etc. of the user, using above-described proposed algorithm and model etc. come real
It is existing.When there is the artificial customer service free time, the music of end block 312 and/or the chat for chatting robot then go to Fig. 2's
The frame 214 of method 200.
However, when in frame 304, when system judges that user is not clear enough to the description of problems/requirements and/or information is inadequate,
Method 300 may be advanced to frame 316.In frame 316, method 300 can determine needs further inquiry user.It is exemplary according to one
And nonlimiting examples, when stopping customer service robot and user into one due to customer service robot and user session have exceeded limitation
Step dialogue is simultaneously assigned user and is determined in not high enough etc. the situation of the confidence level of artificial customer service technical ability group classification, it is believed that
User is not clear enough to the description of problems/requirements and/or information is not enough.In the case, for example, although referring to user
Specific artificial customer service technical ability group is sent to (for example, being responsible for the artificial of difficult miscellaneous artificial customer service technical ability group or another inaccuracy
Customer service technical ability group), but if there is the more time is further to user's proposition further problem and/or request user offer
Information then probably enables a customer to be assigned to more suitable artificial customer service technical ability group, to more accurately be taken
The manual service time needed for business and/or shortening.Therefore, method 300 then returnes to the frame 206 of the method 200 of Fig. 2, thus objective
It takes robot and is further talked with user, for example to obtain further useful information.To which system can in turn again
It is secondary to return to frame 208, to be determined whether that user demand or customer service can be analyzed according to the user information being further collected into
Whether robot and user session have exceeded the automaticmanual customer service worksheet processing for limiting and correspondingly carrying out again.According to another replacement
Property embodiment, can also further talk in customer service robot and user to obtain after further useful information without again
Worksheet processing, but subsequent manual service is carried out using useful information obtained.It, can be in visitor according to another replacement new embodiment
It takes robot and user further talks with to obtain further useful information and carry out selectively worksheet processing later.For example, system
It can determine whether the accuracy of worksheet processing for the first time, and and if only if just worksheet processing again when worksheet processing departure degree is excessive for the first time.For example, if for the first time
User is assigned to close manual service technical ability group by worksheet processing, then can not carry out worksheet processing again;And if worksheet processing is not for the first time
User can be assigned to close manual service technical ability group, then can carry out worksheet processing again.
Further inquiry user is needed determining, the case where confidence level for example to improve the classification of artificial customer service technical ability group
Under, by proposing that further problem and/or request user provide further information to user using busy waiting period, just
It can be effectively by customers assignment to more suitable artificial customer service technical ability group, to more accurately be serviced and/or be shortened institute
Need the manual service time.
Fig. 3 A, Fig. 3 B and Fig. 3 C show the busy waiting scene of personalization accoding to exemplary embodiment.
The scene of Fig. 3 A can be the scene of such as on-line shop's shopping.Information, history, behavior based on user etc., system is true
The elderly men that user A is one's mid-60s is determined, has lived in certain county-level city, received elementary education, it is determined that user A is to purchase
The very dissatisfied desired return of goods of commodity are simultaneously complained.System can determine that the urgency level of the business is for example based on the scheme of the disclosure
Height, and having intense feelings for user is analyzed, it is very discontented.At this point, system can mood and portrait based on user A, correspondingly mention
The busy waiting priority of high user.For example, improving the busy waiting priority of user can include: by user in waiting list
(for example, predetermined ranking or in advance predetermined percentage etc. in advance) in advance, it is preferential to assign experience in corresponding artificial customer service technical ability group
Higher ranked artificial customer service come service the user and/or provide discount coupon or small gift etc. with attempt alleviate user's negative emotions
Deng or any combination thereof.Later, system can mood and portrait based on user A, facilitate use of releiving to recommend or directly play
The classical old song of the mood of family A.
The scene of Fig. 3 B can be the scene for for example handling telephone banking.Information, history based on user, behavior
Deng, system has determined that user B is 30 years old or so unmarried female foreign enterprise white collar, lives in Guangzhou, it is well-educated etc., it determines
The demand of user B is to open up foreign currency current account, and be determined that the time is lunch break etc. at that time.System can be based on this public affairs
The scheme opened determines that the urgency level of the business is for example as low as medium.Due to exchanging by customer service robot and user B,
The demand of user can be determined clearly and without the information for more needing to obtain from user by client machine people, then system
It can be modeled based on the portrait to user B, to recommend some suitable music, such as recent hot music etc. for user B, for
User's B program request.
The scene of Fig. 3 C can be such as APP scene.Information, history, behavior based on user etc., system has determined user
The problem of C is 16 years old or so male student, and user C has substantially been determined is can not to log in original account.System can be based on this public affairs
The scheme opened determine the business urgency level be for example in height.In the frame 208 of such as Fig. 2, because of customer service robot and user C
Talk time or wheel number exceeded limitation, therefore system is automatically into the stage that user is assigned to manual service technical ability group
(for example, frame 210), and the user is assigned to the manual service technical ability group for handling login problem.System enters busy etc. at this time
To the stage.However, user C is not very clear to the description of problem, therefore is due in the dialogue of customer service robot and user C
System can determine needs further inquiry client.At this time system can go to such as Fig. 2 frame 206 and by customer service robot and user into
One step is talked with to obtain more information.
In this example scenario, by the further dialogue of customer service robot and user, more information are got simultaneously
More accurate manual service technical ability group has been reassigned for user.Hereafter, optionally, if artificial customer service is still no-trunk,
System can be for example based on the portrait to user C, and determination will provide chatting service etc. by chat robots for user C.
Fig. 3 A -3C provides the busy example for waiting scene of personalization accoding to exemplary embodiment.By to user's
Music on Demand, the collection and analysis for cutting song and chat information etc., enrich the description that system draws a portrait to user again.These are only
The exemplary and non-limiting example for the scene that the disclosure is applicable in, sole purpose are so that those of ordinary skill in the art
There can be the visual understanding of perception to the technology of the disclosure, and any restriction not constituted to the disclosure.
Such as institute as it can be seen that the present invention can be extensive although being to describe the present invention in conjunction with customer service worksheet processing scene above
User is needed to wait to obtain the scene of service suitable for other.As it is found that be directed to different scenes, can correspondingly design
Different intelligent policy recommendations.
Fig. 4 shows the busy waiting recommendation apparatus 400 of personalization based on machine learning according to the one side of the disclosure
Block diagram.Equipment 400 may include such as human-computer dialogue module 402, user requirements analysis and intention assessment module 404, intelligence group
Single module 406, user and demand portrait modeling module 408, worksheet processing confidence calculations module 410, user emotion analysis module
412, one or more modules in intelligent Service module 414 and intelligent sound box service module 416 etc..
The nonlimiting examples according to exemplary, the executable above frame 202 for combining Fig. 2 of human-computer dialogue module 402,
204, the function of the descriptions such as 206.User requirements analysis and executable above frame 208, the Fig. 3 for combining Fig. 2 of intention assessment module 404
The equal descriptions of frame 306 function.The function of the descriptions such as the executable above frame 210 in conjunction with Fig. 2 of intelligent worksheet processing module 406.User
With the executable above function in conjunction with descriptions such as the frames 302 of frame 202A, Fig. 3 of Fig. 2A of demand portrait modeling module 408.Worksheet processing is set
The executable above function in conjunction with descriptions such as the frames 304,316 of frame 204A, 208A and Fig. 3 of Fig. 2A of reliability computing module 410.
The executable above function in conjunction with descriptions such as the frames 306,308,310 of Fig. 3 of user emotion analysis module 412.Intelligent Service module
The 414 executable above functions in conjunction with descriptions such as the frames 312 and 314 of frame 206A, Fig. 3 of Fig. 2A.Intelligent sound box service module 416
The function of the descriptions such as the executable above frame 314 in conjunction with Fig. 3.
As it is found that above-mentioned function division and appointment is merely exemplary and non-limiting.Implement in other replacements
In example, the function of at least some frames can be merged and be executed by a module and/or the function of at least some frames can be split
To be executed by multiple modules, and can also add/reduce module.
Human-computer dialogue module 402, user requirements analysis and intention assessment module 404, intelligent worksheet processing module 406, Yong Huhe
Demand portrait modeling module 408, worksheet processing confidence calculations module 410, user emotion analysis module 412, intelligent Service module
One or more modules in 414 and intelligent sound box service module 416 etc. can be realized by various modes.According to
One exemplary and non-limiting embodiment, above-mentioned module can be realized by software.In such embodiments, above-mentioned module can quilt
It stores in memory and is executed by a processor to realize its corresponding function.
According to other embodiments, above-mentioned module also may be implemented as hardware, such as ASIC, field programmable gate array
(FPGA), discrete logic unit etc..Above-mentioned module also may be implemented as software, be such as stored in memory 404 and by
Processor 402 is executed to realize corresponding function.Above-mentioned module can also be implemented as the various combinations of hardware and software.These are
Within the scope of the present disclosure.
The new busy waiting strategy that the scheme of the disclosure is proposed can both have been enriched to the understanding of customer problem/demand,
Worksheet processing accuracy rate can be promoted again.Moreover, the scheme of the disclosure can also promote user satisfaction.
It will be recognized by one of ordinary skill in the art that the beneficial effect of the disclosure is not by any single embodiment Lai all real
It is existing.Various combinations, modification and replacement are that those of ordinary skill in the art are illustrated on the basis of the disclosure.
In addition, term "or" is intended to indicate that inclusive "or" and nonexcludability "or".That is, unless otherwise specified or from upper and lower
Text can be clearly seen, otherwise phrase " X " use " A " or " B " be intended to indicate that it is any naturally can and arrangement.That is, phrase " X " is adopted
Met with " A " or " B " by example any in following instance: X uses A;X uses B;Or X uses both A and B.Belong to
" connection " can indicate identical meanings with " coupling ", indicate the electrical connection of two devices.In addition, disclosure and the accompanying claims book
Used in the article " one " and " certain " be generally to be understood as indicating " one or more ", unless stated otherwise or can be from upper
It is hereinafter apparent from and refers to singular.
Various aspects or feature by by may include several equipment, component, module, and the like system in the form of be in
It is existing.It should be understood that and understand, various systems may include optional equipment, component, module etc., and/or can not include in conjunction with attached drawing
Armamentarium, component, module for being discussed etc..Also the combination of these methods can be used.
It can be with general in conjunction with various illustrative logicals, logical block, module and the circuit that presently disclosed embodiment describes
Processor, digital signal processor (DSP), specific integrated circuit (ASIC), field programmable gate array (FPGA) or it is other can
Programmed logic device, discrete door or transistor logic, discrete hardware component or its be designed to carry out function described herein
Any combination realize or execute.General processor can be microprocessor, but in alternative, and processor, which can be, appoints
What conventional processor, controller, microcontroller or state machine.Processor is also implemented as calculating the combination of equipment, example
As DSP and the combination of microprocessor, multi-microprocessor, the one or more microprocessors cooperateed with DSP core or it is any its
Its such configuration.In addition, at least one processor may include may act on execute one or more steps described above and/or
One or more modules of movement.For example, above can be by processor and being coupled in conjunction with the embodiment of each method description
The memory of processor realizes, wherein the processor can be configured to execute any step of aforementioned any method or its is any
Combination.
In addition, the method that describes in conjunction with aspect disclosed herein or the step of algorithm and/or movement can be directly hard
Implement in part, in the software module executed by processor or in combination of the two.For example, combining each method above
The embodiment of description can realize by being stored with the computer-readable medium of computer program code, wherein the computer journey
Sequence code executes any step or any combination thereof of aforementioned any method when being executed by processor/computer.
The element of the various aspects described in the whole text in the disclosure is that those of ordinary skill in the art are currently or hereafter known
Equivalent scheme in all structures and functionally is clearly included in this by citation, and is intended to be intended to be encompassed by the claims.
In addition, any content disclosed herein is all not intended to contribute to the public --- it is no matter such open whether in claim
It is explicitly recited in book.