CN109274845A - Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically - Google Patents

Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically Download PDF

Info

Publication number
CN109274845A
CN109274845A CN201811010990.1A CN201811010990A CN109274845A CN 109274845 A CN109274845 A CN 109274845A CN 201811010990 A CN201811010990 A CN 201811010990A CN 109274845 A CN109274845 A CN 109274845A
Authority
CN
China
Prior art keywords
target
data
voice
outgoing call
return visit
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811010990.1A
Other languages
Chinese (zh)
Inventor
黄锦伦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Priority to CN201811010990.1A priority Critical patent/CN109274845A/en
Publication of CN109274845A publication Critical patent/CN109274845A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M3/00—Automatic or semi-automatic exchanges
    • H04M3/42—Systems providing special services or facilities to subscribers
    • H04M3/50—Centralised arrangements for answering calls; Centralised arrangements for recording messages for absent or busy subscribers ; Centralised arrangements for recording messages
    • H04M3/527—Centralised call answering arrangements not requiring operator intervention
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L17/00—Speaker identification or verification
    • G10L17/04—Training, enrolment or model building
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L17/00—Speaker identification or verification
    • G10L17/22—Interactive procedures; Man-machine interfaces
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M3/00—Automatic or semi-automatic exchanges
    • H04M3/42—Systems providing special services or facilities to subscribers

Abstract

The present invention discloses a kind of intelligent sound and pays a return visit method, apparatus, computer equipment and storage medium automatically, applies in field of speech recognition.This method comprises: obtaining voice callback access request, issue database is inquired based on the target channel mark in voice callback access request, target voice is obtained and pays a return visit problem and target answer data；Based on target channel mark inquiring customer information database, automatic speech outgoing call tables of data is obtained；Automatic speech outgoing call tables of data and target voice are paid a return visit into problem and are pushed to telephony platform, so that telephony platform is based on return visit voice data corresponding with target voice return visit problem and automatic speech outgoing call tables of data carries out voice return visit, the outgoing call recording data that telephony platform returns is obtained；Text translation is carried out to outgoing call recording data using target voice static state decoding network, obtains outgoing call text data；Outgoing call text data and target answer data are analyzed, automatic speech outgoing call result is obtained.This method can realize that voice is paid a return visit and analysis is paid a return visit as a result, improving efficiency and reducing cost.

Description

Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically
Technical field
The present invention relates to technical field of voice recognition more particularly to a kind of intelligent sound to pay a return visit method, apparatus automatically, calculates Machine equipment and storage medium.
Background technique
In the financial business such as bank, security, insurance or other business procedures, it is usually provided with client and pays a return visit process, To understand the satisfaction and improvement idea of client, to facilitate carry out business improvement.Existing customer pays a return visit process mainly by artificial It attends a banquet and carries out telephonic communication with client, process needs higher human cost, and pays a return visit timeliness and pay a return visit inefficient.And And after operator attendance and client carry out results of telephone call back, it need to be carried out by return visit result of the operator attendance to each client Statistical analysis, time-consuming for process, low efficiency and at high cost.
Summary of the invention
The embodiment of the present invention provides a kind of intelligent sound and pays a return visit method, apparatus, computer equipment and storage medium automatically, with It solves the problems, such as that current manual attends a banquet and carries out client's return visit and the process cost analyzed is higher and efficiency is lower.
A kind of intelligent sound pays a return visit method automatically, comprising:
Voice callback access request is obtained, the voice callback access request includes target channel mark；
Issue database is inquired based on the target channel mark, obtains target corresponding with the target channel mark Voice pays a return visit problem and target answer data corresponding with target voice return visit problem；
Based on the target channel mark inquiring customer information database, obtain corresponding with the target channel mark Automatic speech outgoing call tables of data；
The automatic speech outgoing call tables of data and the target voice are paid a return visit into problem and are pushed to telephony platform, so that described Telephony platform is based on paying a return visit the corresponding return visit voice data of problem and the automatic speech outgoing call tables of data with the target voice Voice return visit is carried out, the outgoing call recording number corresponding with target voice return visit problem that the telephony platform returns is obtained According to；
Text translation is carried out to the outgoing call recording data using target voice static state decoding network, is obtained and the target Voice pays a return visit the corresponding outgoing call text data of problem；
Divide with the target voice return visit corresponding outgoing call text data of problem and the target answer data Analysis obtains automatic speech outgoing call result.
A kind of automatic playback appliances of intelligent sound, comprising:
Voice callback access request obtains module, and for obtaining voice callback access request, the voice callback access request includes target canal Road mark；
Target return visit problem obtains module, for inquiring issue database, acquisition and institute based on the target channel mark It states the corresponding target voice return visit problem of target channel mark and target corresponding with target voice return visit problem is answered Case data；
Outgoing call tables of data obtain module, for be based on the target channel mark inquiring customer information database, obtain with The corresponding automatic speech outgoing call tables of data of the target channel mark；
Data-pushing processing module is pushed away for the automatic speech outgoing call tables of data and the target voice to be paid a return visit problem Telephony platform is given, so that the telephony platform is based on the return visit voice data corresponding with target voice return visit problem Voice return visit is carried out with automatic speech outgoing call tables of data, obtain the telephony platform return pays a return visit problem with the target voice Corresponding outgoing call recording data；
Outgoing call text data obtain module, for using target voice static state decoding network to the outgoing call recording data into It composes a piece of writing this translation, obtains outgoing call text data corresponding with target voice return visit problem；
Voice outgoing call result obtains module, for outgoing call text data corresponding with target voice return visit problem It is analyzed with the target answer data, obtains automatic speech outgoing call result.
A kind of computer equipment, including memory, processor and storage are in the memory and can be in the processing The computer program run on device, the processor realize the above-mentioned intelligent sound side of return visit automatically when executing the computer program The step of method.
A kind of computer readable storage medium, the computer-readable recording medium storage have computer program, the meter Calculation machine program realizes the step of above-mentioned intelligent sound pays a return visit method automatically when being executed by processor.
Above-mentioned intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically, based in voice callback access request Target channel mark inquire corresponding database, obtain target voice respectively and pay a return visit problem and automatic speech outgoing call tables of data, Acquisition process is simple and convenient.Then, automatic speech outgoing call tables of data and target voice are paid a return visit into problem and is pushed to telephony platform, with Telephony platform is set to pay a return visit problem to the corresponding client of customer data in the automatic speech outgoing call tables of data according to the target voice Voice return visit is carried out, to obtain the outgoing call recording data of telephony platform feedback, to improve the efficiency of voice return visit and reduce voice The cost of return visit.Text translation is carried out to outgoing call recording data using target voice static state decoding network, accuracy rate is higher and knows Other fast speed.Finally, being divided based on the corresponding outgoing call text data of target voice return visit problem and target answer data Analysis, can the more intuitive automatic speech outgoing call of quick obtaining as a result, with improve voice pay a return visit analysis efficiency and advantageously reduce language Sound pays a return visit the cost of analysis.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below by institute in the description to the embodiment of the present invention Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings Obtain other attached drawings.
Fig. 1 is the application environment schematic diagram that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 2 is the flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 3 is another flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 4 is another flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 5 is another flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 6 is another flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 7 is another flow chart that intelligent sound pays a return visit method automatically in one embodiment of the invention；
Fig. 8 is a schematic diagram of the automatic playback appliances of intelligent sound in one embodiment of the invention；
Fig. 9 is a schematic diagram of computer equipment in one embodiment of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts Example, shall fall within the protection scope of the present invention.
Intelligent sound provided in an embodiment of the present invention pays a return visit method automatically, and the method for paying a return visit can be using such as automatically for the intelligent sound In application environment shown in FIG. 1.Specifically, the intelligent sound automatically apply in voice return visit system by the method for paying a return visit, the voice Return visit system includes client and server as shown in Figure 1, and client is communicated with server by network, for realizing The function that automatic speech return visit is carried out to client and is analyzed.Wherein, client is also known as user terminal, refers to opposite with server It answers, provides the program of local service for client.Client it is mountable but be not limited to various personal computers, laptop, On smart phone, tablet computer and portable wearable device.Server can use the either multiple services of independent server The server cluster of device composition is realized.
In one embodiment, it as shown in Fig. 2, provide a kind of intelligent sound pays a return visit method automatically, is applied in this way in Fig. 1 In server for be illustrated, include the following steps:
S201: voice callback access request is obtained, voice callback access request includes target channel mark.
Wherein, to be user pay a return visit function by what user end to server was sent for realizing voice to voice callback access request Request.Target channel mark is for the corresponding channel mark of unique identification target service channel, which refers to this Channel used by client's transacting business corresponding to secondary voice callback access request.By taking loan transaction as an example, client is handling loan In business procedure, corresponding business channel includes that new loan channel and Fei Xin borrow channel, therefore, entrained by the voice callback access request Target channel mark can borrow channel mark or non-new loan channel mark to be new.
Further, which can also include that target service identifies, and target service mark is for only One identification needs to carry out the mark of the target service of voice return visit.Target channel mark and target industry are carried in voice callback access request Business mark, so that voice return visit system can be according to the voice callback access request, to target corresponding with the target channel mark Business channel handles the client for identifying corresponding business with target service and carries out voice return visit.For example, the voice callback access request In target service be identified as loan transaction mark, target channel mark be it is new borrow channel mark, then voice pay a return visit system according to The voice callback access request handles the client of loan transaction in a manner of channel and carries out voice return visit, to improve client's return visit to newly to borrow Efficiency and reduce and artificial pay a return visit cost.
S202: issue database is inquired based on target channel mark, obtains target language corresponding with target channel mark Sound pays a return visit problem and target answer data corresponding with target voice return visit problem.
Wherein, issue database is the database that problem and corresponding answer data are paid a return visit for storaged voice.In problem In database, it is previously stored with multiple voices and pays a return visit problem, it is associated with channel mark that each voice pays a return visit problem.
It is that voice corresponding with the target channel mark return visit stored in issue database is asked that target voice, which pays a return visit problem, Topic.Target answer data is pre-stored corresponding with target voice return visit problem for analyzing return visit in issue database As a result data.Specifically, server inquires issue database according to the target channel mark carried in voice callback access request, from Target voice corresponding with target channel mark is obtained in the issue database pays a return visit problem and corresponding target answer number According to.In loan transaction, target channel mark can borrow channel mark or non-new loan channel mark to be new, correspondingly, Voice corresponding with loan transaction mark with what is stored in issue database pays a return visit problem and specifically includes and newly borrow channel mark Or the corresponding voice of non-new loan channel mark pays a return visit problem.
Further, target voice return visit problem can be associated with target service mark and target channel mark Voice pays a return visit problem, to determine that corresponding voice pays a return visit problem according to target service and target channel, so that target voice The specific aim of return visit problem is stronger, so that carrying out the automatic speech outgoing call result that gets of analysis more according to the return visit result of client The opinion that can reflect client is more advantageous to business improvement.In the present embodiment, server mesh according to entrained by voice callback access request Mark service identification and inquire issue database, obtain corresponding with target service mark voice return visit problem, then from target industry Business identifies in corresponding voice return visit problem and pays a return visit problem according to the corresponding target voice of target channel mark inquiry acquisition With target answer data.
For example, it includes the following: A01- to guarantee that your pacify at information that voice corresponding with non-new loan channel mark, which pays a return visit problem, It needs to check identity information with you entirely.Does is the birthday that may I ask you * * month * day? (time system enters for, and the moon, day client return It is multiple).Does is it how many ten thousand that A02-, which may I ask your amount of money of loan,? (answering questions ten thousand (containing ten thousand) numbers above to calculate unanimously).If A03- A02 verification is lost Does is lose and trigger the problem: may I ask your release date of loan * * month * day? if (answer records inconsistent, turn verification mistake with system Lose conclusion).Does is does is A04- may I ask that you manage it to me, and this for the new loan service that you provide satisfied, general or unsatisfied with? If (answering dissatisfied triggering problem) which aspect enables you dissatisfied? (recording dissatisfied reason).A05- be may I ask in this loan During money is handled, if there is bank clerk to charge to you? (if answer have triggering should) have collected you how much? (note The expense that record client says, triggering the problem of you feed back, we have helped you to register and have fed back relevant departments, thank to your cooperation.).
Relatively, with new to borrow the corresponding voice of channel mark to pay a return visit problem include the following: B01- to guarantee your information Security needs check identity information with you.Does is the birthday that may I ask you * * month * day? (time system enters for, the moon, day client It replys).Does is it how many ten thousand that B02-, which may I ask your amount of money of loan,? (answering questions ten thousand (containing ten thousand) numbers above to calculate unanimously).If B03- B02 is checked Does is unsuccessfully trigger the problem: may I ask your release date of loan * * month * day? if (answer records inconsistent, turn verification with system Failure conclusion).B04- be may I ask during handling loan, and whether telemarketing personnel have through number in addition to " 4008003333 " Does code contact loan matters with you? if (answer has the problem) may I ask telephone number is how many? (record telephone number).B05- may I ask Does is your your new one satisfaction for borrowing service for providing of my behavior to this, satisfied, general or dissatisfied? if (answering not It is satisfied to trigger the problem) do which aspect enable you dissatisfied? (recording dissatisfied reason).B06-, which may I ask, handles process in this loan In, if there is bank clerk to charge to you? (if answer have triggering should) have collected you how much? (record client says Expense, triggering the problem of you feed back, we have helped you to register and have fed back relevant departments, thank to your cooperation.)
In the present embodiment, each voice in issue database pays a return visit the corresponding question number of problem, the question number For unique identification, its corresponding voice pays a return visit problem, carries out return visit interpretation of result so as to subsequent.Correspondingly, it is paid a return visit with the voice The corresponding answer data of problem also includes identical question number.Target answer data is pre-stored in issue database One answer data specifically includes question number, at least two option names and option corresponding with each option names and compiles Number.Wherein, option number is the number of its corresponding option names for unique identification.For example, being QT01 for question number Voice pay a return visit problem, the option names in answer data include that satisfied, dissatisfied, general and client mismatches this four choosings , respectively correspond this four option numbers of QP01, QP02, QP03 and QP04.
S203: being based on target channel mark inquiring customer information database, and acquisition is corresponding with target channel mark certainly Dynamic voice outgoing call tables of data.
Wherein, customer information database is the database for storing customer data.The customer information database is for depositing Store up the customer data that different business systems upload.Each customer data include customer ID, customer name, the date of birth, gender, Work Telephone, phone number, service identification and channel mark can also include business datum corresponding with service identification.It should Customer ID can be identification card number or user account number etc. can be with the mark of unique identification client identity.Service identification is to be used for Identify the mark for the business that the client is handled.Business datum is that client handles shape in business procedure corresponding with service identification At data.By taking loan transaction as an example, which includes the amount of the loan, type of providing a loan and handles the data such as customer manager.
Automatic speech outgoing call tables of data is the tables of data for storing the customer data for needing to carry out voice return visit, the data Table needs to be issued to telephony platform, so that telephony platform is corresponding to each customer data in the automatic speech outgoing call database Client carries out voice return visit.
In the present embodiment, server is based on target channel mark inquiring customer information database, from the customer profile data Customer data corresponding with target channel mark is obtained in library, and automatic speech outgoing call tables of data is formed based on the customer data, So that the customer data in automatic speech outgoing call tables of data is associated with target channel mark, so that telephony platform can be to this certainly Customer data in dynamic voice outgoing call tables of data carries out batch processing, and asks for target voice return visit in batch process Topic carries out voice return visit, improves the specific aim that voice is paid a return visit.
Further, if voice callback access request also carries target service mark, server is also needed based on target service Mark and target channel mark inquiring customer information database, from the customer information database obtain with target service mark and The corresponding customer data of target channel mark forms automatic speech outgoing call tables of data based on the customer data, so that this is automatic Customer data in voice outgoing call tables of data is associated with target channel mark and target service mark, to improve telephony platform Carry out the specific aim of voice return visit.
S204: paying a return visit problem for automatic speech outgoing call tables of data and target voice and be pushed to telephony platform, so that phone is flat Stylobate carries out voice return visit in return visit voice data corresponding with target voice return visit problem and automatic speech outgoing call tables of data, obtains Take the outgoing call recording data corresponding with target voice return visit problem that telephony platform returns.
Specifically, kettle tool is previously provided in server, Kettle tool is used for automatic speech outgoing call data Table and target voice pay a return visit problem and are pushed to telephony platform, so that telephony platform is asked based on the target voice return visit of server push Corresponding return visit voice data is inscribed, voice return visit is carried out to the corresponding client of customer data in automatic speech outgoing call tables of data. Kettle tool has GUI graphical interfaces, supports multithreading lot size scheduling, processing speed is fast and has perfect monitoring log etc. Advantage.
Specifically, server by automatic speech outgoing call tables of data and target voice pay a return visit problem be pushed to telephony platform it Afterwards, so that telephony platform be based on target voice pay a return visit the corresponding return visit voice data of problem and automatic speech outgoing call tables of data into Row voice is paid a return visit, i.e., so that telephony platform pays a return visit corresponding return visit voice data according to the target voice, to automatic speech outgoing call The corresponding client of customer data in tables of data carries out call-on back by phone and records, to obtain outgoing call recording data.Phone is flat Platform after the corresponding outgoing call recording data of all customer datas, can record the outgoing call in obtaining automatic speech outgoing call tables of data Data return to server, so that server obtains the outgoing call corresponding with target voice return visit problem that telephony platform returns and records Sound data, to be analyzed and processed to the outgoing call recording data.
S205: text translation, acquisition and target language are carried out to outgoing call recording data using target voice static state decoding network Sound pays a return visit the corresponding outgoing call text data of problem.
Target voice static state decoding network is in advance using the training text data of specific area and corresponding registration voice Data carry out the static decoding network of content of text in the voice for identification got after model training.The training text data Specially pay a return visit the corresponding text data of problem with target voice, for example, server by network acquire user's upload with Target voice pays a return visit the corresponding text data of problem, and this article notebook data is stored in training data corpus, in model When training, the text data got from the training data corpus is training text data.Training voice data is to adopt The voice data accessed by voice sampling instrument acquisition user's reading training text data.Since target voice static state decodes Network is that the training text data based on specific area are trained acquired static decoding network, so that it is to the specific neck When the outgoing call recording data in domain is identified, specific aim is stronger, so that decoding accuracy rate is higher.Due to static decoding network Search space is all unfolded, therefore it when carrying out text translation, decoding speed is very fast, so as to quick obtaining outgoing call text Notebook data.In the present embodiment, text translation is carried out to outgoing call recording data using target voice static state decoding network, can quickly be obtained Take the higher outgoing call text data of recognition accuracy.The outgoing call text data is using target voice static state decoding network to outgoing call After recording data is identified, the data existing in the form of text that get.
S206: analyzing outgoing call text data corresponding with target voice return visit problem and target answer data, Obtain automatic speech outgoing call result.
Specifically, server in obtaining automatic speech outgoing call tables of data the corresponding client feedback of each customer data it is outer It is after the outgoing call text data for exhaling recording data to be converted into, the outgoing call text data is corresponding with target voice return visit problem Target answer data compares, to obtain the feedback opinion for being convenient for system statistics；Again based in automatic speech outgoing call tables of data The feedback opinion of the corresponding client of all customer datas, forms automatic speech outgoing call result.Specifically, the automatic speech outgoing call knot Fruit tabular form shows that each client pays a return visit target voice the feedback opinion of problem, so that the automatic speech outgoing call result can be straight Feedback opinion of the client to the business that related service is handled in display by the corresponding target service channel of target channel mark is seen, Business improvement is carried out so that operation system is based on the feedback opinion.
Intelligent sound provided by the present embodiment is automatically in return visit method, based on the target channel mark in voice callback access request Know and inquire corresponding database, obtains target voice return visit problem respectively and automatic speech outgoing call tables of data, acquisition process are simple It is convenient.Then, automatic speech outgoing call tables of data and target voice are paid a return visit into problem and are pushed to telephony platform so that telephony platform according to Problem is paid a return visit according to the target voice, and voice return visit is carried out to the corresponding client of customer data in the automatic speech outgoing call tables of data, To obtain the outgoing call recording data of telephony platform feedback, to improve the efficiency of voice return visit and reduce the cost of voice return visit.It adopts Text translation is carried out to outgoing call recording data with target voice static state decoding network, accuracy rate is higher and recognition speed is very fast.Most Afterwards, the corresponding outgoing call text data of problem is paid a return visit based on target voice and target answer data is analyzed, it can quick obtaining More intuitive automatic speech outgoing call as a result, with improve voice pay a return visit analysis efficiency and advantageously reduce voice pay a return visit analysis at This, it helps operation system is based on the automatic speech outgoing call result and carries out business improvement.
In one embodiment, as shown in figure 3, the intelligent sound returns automatically before the step of obtaining voice callback access request Visit method further include:
S301: obtaining outgoing call list data, and outgoing call list data includes at least one newly-increased customer data, increases client's number newly According to including customer ID.
Wherein, outgoing call list data is the list data that operation system is sent to that voice pays a return visit system, the outgoing call name odd number According to the list data by needing the customer data for carrying out voice return visit to form.Newly-increased customer data refers to this outgoing call name odd number Customer data in.Each newly-increased customer data include customer ID, customer name, the date of birth, gender, Work Telephone, Phone number, service identification and channel mark can also include business datum corresponding with service identification.The customer ID can To be that identification card number and user account number etc. can be with the marks of unique identification client identity.
It, will be with default when loan transaction system pays a return visit system transmission outgoing call list data to voice by taking loan transaction as an example Upload mode uploads new loan channel referral list or non-both outgoing call list datas of new loan channel referral list.Wherein, new to borrow The difference that channel referral list and Fei Xin borrow channel referral list is, new to borrow each newly-increased customer data in channel referral list This corresponding " new to borrow " new loan channel mark of channel mark field, rather than newly borrow each newly-increased client's number in channel referral list According to this corresponding " non-new " non-new loan channel mark of channel mark field.
S302: being based on customer ID inquiring customer information database, judges that customer information database whether there is and client Identify corresponding existing customer data.
Customer information database is the database for storing customer data, is particularly used for storing existing customer data Database, each corresponding customer ID of existing customer data.In the present embodiment, server receive outgoing call list data it Afterwards, the customer ID of the customer ID of each newly-increased customer data in the outgoing call list data and existing customer data is carried out Comparison, to determine in customer information database with the presence or absence of existing client's number corresponding with the customer ID of newly-increased customer data According to so that it is determined that whether newly-increased customer data has stored in customer information database, so as to according to existing customer data Task status carries out voice outgoing call processing.
S303: if existing customer data corresponding with customer ID is not present in customer information database, by newly-increased visitor User data is stored in customer information database, and configuring the newly-increased corresponding task status of customer data is new task state.
In the present embodiment, task status includes but is not limited to new task state, automatic speech outgoing call status of fail and automatic Voice outgoing call complain state these three.Wherein, new task state, which refers to, carries out automatic speech return visit by telephony platform not yet The state of processing.Automatic speech outgoing call status of fail refers to that carrying out automatic speech by telephony platform pays a return visit processing and return visit It as a result is the state of failure.Automatic speech outgoing call complains state to refer to and carries out automatic speech return visit processing by telephony platform And paying a return visit result is the state complained.
Specifically, if existing visitor corresponding with the customer ID of newly-increased customer data is not present in customer information database User data illustrates that the corresponding client of newly-increased customer data in the outgoing call list data is new client, at this point, by newly-increased client's number According to being stored in customer information database, configuring the newly-increased corresponding task status of customer data is new task state, so as to root The configuration of automatic speech outgoing call tables of data is carried out, according to the new task state to realize Classification Management.
S304: if there is existing customer data corresponding with customer ID in customer information database, using newly-increased Customer data updates existing customer data, and retains the existing corresponding task status of customer data.
Specifically, if there is existing client corresponding with the customer ID of newly-increased customer data in customer information database Data illustrate that the corresponding client of newly-increased customer data in the outgoing call list data is existing client, for existing client, Voice pays a return visit system may carry out voice return visit to it, it is also possible to voice return visit is not carried out, therefore, the existing customer data Corresponding task status may be in new task state, automatic speech outgoing call status of fail and automatic speech outgoing call complaint state It is any, therefore, the existing corresponding task status of customer data need to be retained, carried out accordingly so as to subsequent according to its task status Processing, rather than directly as new work be state processing, can help to improve subsequent processing specific aim.It is understood that Ground is uploaded to voice due to having customer data and newly-increased customer data and pays a return visit the process of system there are the time difference, at this time client Data may change (Work Telephone or other information in such as customer data change) or due to data inputting Reason causes the two data inconsistent, at this point, newly-increased customer data need to be updated to existing customer data, to guarantee customer information number According to the real effectiveness of the customer data in library.
Intelligent sound provided by the present embodiment is automatically in return visit method, according to newly-increased client each in outgoing call list data The customer ID inquiring customer information database of data, to determine whether there is existing client's number corresponding with the customer ID According to the existing customer data of presence and there is no two kinds of judging results of existing customer data to be respectively processed, so that this is newly-increased Customer data configures different task statuses, is handled respectively so as to subsequent according to different task statuses, to improve subsequent voice Automatically the specific aim and validity paid a return visit.
In one embodiment, as shown in figure 4, being based on target channel mark inquiring customer information database, acquisition and target The corresponding automatic speech outgoing call tables of data of channel mark, comprising:
S401: voice outgoing call task is obtained, voice outgoing call task includes target channel mark and goal task state.
Wherein, voice outgoing call task refers to that client is sent to being formed outside automatic speech for trigger the server for server Exhale the task of tables of data.Target channel mark is for the corresponding channel mark of unique identification target service channel, in loan industry In business, target channel mark can borrow channel mark or non-new loan channel mark to be new.Goal task state refers to this language The task status selected in sound outgoing call task, including but not limited to new task state, automatic speech outgoing call status of fail and automatic Voice outgoing call complain state these three.
In the present embodiment, on the client operation page that voice pays a return visit system, it is configured with optional channel and task status Corresponding drop-down list.It is corresponding with selection target channel, such as loan transaction that user can click the drop-down list of the optional channel Any of " new borrow " and " non-new " as target channel mark；The corresponding drop-down list of the task status is clicked, with choosing Any of new task state, automatic speech outgoing call status of fail and automatic speech outgoing call complaint state are selected as goal task State, to input voice outgoing call task, so that server obtains voice outgoing call task.
S402: it is based on target channel mark and goal task status inquiry customer information database, obtains effective client's number According to.
Specifically, server is after receiving voice outgoing call task, according to target channel mark therein and goal task Status inquiry customer information database, to be obtained and target channel mark and goal task state phase from customer information database Matched effective customer data, effective customer data are to meet client corresponding to target channel mark and goal task state Data.It is to be appreciated that effective customer data is the customer data that can be handed down to telephony platform and carry out voice return visit.
S403: being based on target channel mark and goal task state, determines that optional outgoing call issues mode, and according to optional outer The mode of issuing is exhaled to determine that target outgoing call issues mode.
Optional outgoing call issue mode be by this voice outgoing call task target channel mark and goal task state it is true The fixed outgoing call for user's selection issues mode, specifically includes that artificial outgoing call issues mode and automatic outer call issues mode two Kind.Target outgoing call issues mode and refers to that the outgoing call selected in this voice outgoing call task issues mode.Due to target channel mark It is one of optional channel of N kind, goal task state is one of M kind task status, and the two combines outside the voice to be formed Exhaling task includes N*M kind, and the optional outgoing call in the present embodiment issues mode and only includes artificial outgoing call and issues mode and automatic outer Two kinds of the mode of issuing are exhaled, therefore server need to be previously stored with mission dispatching relation table, which is stored with often The corresponding optional outgoing call of the syntagmatic of one target channel mark and goal task state issues mode, so that user selects It selects.It is to be appreciated that if optional outgoing call issues mode and is defaulted as that artificial outgoing call issues mode and automatic outer call issues in mode One, then the optional outgoing call of default is issued into mode and be determined as target outgoing call and issue mode；If optional outgoing call issues mode Artificial outgoing call issues mode and automatic outer call issues both in mode, then it is that user passes through client that target outgoing call, which issues mode, The optional outgoing call determined after being selected issues mode.
In the present embodiment, server determines its corresponding optional outgoing call according to target channel mark and goal task state Mode is issued, and the optional outgoing call is issued into mode and is shown on the client operation page, so that user selects, Yong Huke It is issued in mode from optional outgoing call and selects one of them, it, will be outside automatic speech with determination in a manner of being determined as target outgoing call and issue Tables of data is exhaled to be pushed to the mode of telephony platform.
S404: when it is that automatic outer call issues mode that target outgoing call, which issues mode, target push date and destination name are obtained Odd number is measured, if the target push date meets default push rule, selection and the consistent effective customer data of destination name odd number amount, As automatic speech outgoing call tables of data corresponding with target channel mark.
The target push date refers to the push date that this voice outgoing call task determines.Destination name odd number amount refers to this language The quantity of the customer data of sound outgoing call required by task.In advance push rule be for assess target push the date whether meet it is default The rule of standard is specifically as follows in preset time range of the target push date after the current time in system.
In the present embodiment, if it is that automatic outer call issues mode that target outgoing call, which issues mode, needs to obtain user and pass through visitor It is the target push date and destination name odd number amount that family end uploads, how much objective to telephony platform automatic push to determine when User data.Server is first sentenced after receiving the target push date and destination name odd number amount that user is uploaded by client Whether the target of breaking push date meets default push rule, if not meeting, change date information is sent to client, to mention The user that wakes up carries out target push date modification；If meeting, selected from customer information database consistent with destination name odd number amount Effective customer data, as automatic speech outgoing call tables of data corresponding with target channel mark.
It is to be appreciated that obtaining automatic speech in the case where delivering under target outgoing call issues mode as automatic outer call After outgoing call tables of data, server real-time acquisition system current time, and when the current time in system is that target pushes the date, it will The automatic speech outgoing call tables of data is pushed to telephony platform, so that telephony platform is to client's number in automatic speech outgoing call tables of data Automatic speech return visit is carried out according to corresponding client.
S405: when it is that artificial outgoing call issues mode that target outgoing call, which issues mode, destination name odd number amount, selection and mesh are obtained The consistent effective customer data of entitling odd number amount, as automatic speech outgoing call tables of data corresponding with target channel mark.
In the present embodiment, if it is that artificial outgoing call issues mode that target outgoing call, which issues mode, server receives user and passes through After the destination name odd number amount that client uploads, selection and the consistent effective visitor of destination name odd number amount from customer information database User data, as automatic speech outgoing call tables of data corresponding with target channel mark.
It is to be appreciated that server is obtaining automatic language when it is that artificial outgoing call issues mode that target outgoing call, which issues mode, After sound outgoing call tables of data, which can be pushed to telephony platform, so that telephony platform is to automatic The corresponding client of customer data in voice outgoing call tables of data carries out automatic speech return visit.
Intelligent sound provided by the present embodiment is automatically in return visit method, and elder generation is according to the target channel in voice outgoing call task Mark and goal task state determine effective customer data, are selected with realizing according to user, determine this voice outgoing call task institute Corresponding effective customer data, it is with strong points.Then, it is determined under optional outgoing call according to target channel mark and goal task state Originating party formula, corresponding target outgoing call issues mode for selection by the user, and the difference of mode is issued according to target outgoing call, determines it Required destination name odd number amount and whether need to be arranged the target push date, in a manner of meeting different target outgoing call and issue needed for Pushed information, to form automatic speech outgoing call tables of data.
In one embodiment, as shown in figure 5, automatic speech outgoing call tables of data and target voice, which are paid a return visit problem, is pushed to electricity Platform is talked about, so that telephony platform is based on return visit voice data corresponding with target voice return visit problem and automatic speech outgoing call data Table carries out voice return visit, comprising:
S501: inquiring issue database based on target channel mark, obtains target voice and pays a return visit the corresponding push shape of problem State.
Target voice pays a return visit the corresponding push state of problem and refers to that target voice pays a return visit whether problem is once pushed to electricity Talk about the state of platform.In the present embodiment, the push state is including having pushed state and not pushed state.In issue database, It is stored with multiple voices and pays a return visit problem and its corresponding answer data, be also stored with the push state that each voice pays a return visit problem, It is associated with channel mark that each voice pays a return visit problem.Specifically, server is based on target channel mark and inquires issue database, Target voice corresponding with target channel mark is obtained from issue database pays a return visit problem and its corresponding push state, with Continue after an action of the bowels and is handled.
S502: if push state is to have pushed state, automatic speech outgoing call tables of data and target voice are paid a return visit into problem Corresponding question number is pushed to telephony platform, so that telephony platform is based on the corresponding return visit voice data of question number and automatically Voice outgoing call tables of data carries out voice return visit.
In the present embodiment, if the push state that target voice pays a return visit problem is to have pushed state, illustrate that the target voice returns Access topic has been pushed to telephony platform before this, and telephony platform is stored with the target voice and pays a return visit problem and corresponding time Voice data is visited, and the target voice pays a return visit problem and pays a return visit the problem of voice data pays a return visit problem with target voice and numbers It is associated.Therefore, server is when the push state that target voice pays a return visit problem is to have pushed state, by automatic speech outgoing call number The corresponding question number of problem is paid a return visit according to table and target voice and is pushed to telephony platform, so that telephony platform is based on the question number Corresponding return visit voice data and automatic speech outgoing call tables of data carry out voice return visit, to reduce volume of transmitted data, improve data Pushing efficiency.Server is based on paying a return visit voice data and automatic speech outgoing call tables of data carries out voice return visit, and process is not necessarily to people Work intervention can effectively improve the efficiency of voice return visit, reduce the cost of labor that voice is paid a return visit.
S503: if push state is not push state, automatic speech outgoing call tables of data and target voice are paid a return visit into problem It is pushed to telephony platform, so that telephony platform is based on target voice and pays a return visit the corresponding return visit voice data of problem acquisition, and is based on It pays a return visit voice data and automatic speech outgoing call tables of data carries out voice return visit.
In the present embodiment, if the push state that target voice pays a return visit problem is not push state, illustrate that the target voice returns Access topic is not pushed to telephony platform before this, and telephony platform is not stored with target voice and pays a return visit problem and corresponding Pay a return visit voice data.At this point, automatic speech outgoing call tables of data and target return visit phonetic problem need to be pushed to phone and put down by server Platform is paid a return visit problem and is collected and corresponding pay a return visit voice data (i.e. manual read should so that telephony platform is first based on the target voice Target voice pays a return visit problem and simultaneously records, and so that it is paid a return visit problem with target voice associated), be then based on the return visit voice data with Automatic speech outgoing call tables of data carries out voice return visit, and process is not necessarily to manual intervention, can effectively improve the efficiency of voice return visit, drop The cost of labor that sound of speaking in a low voice is paid a return visit.
In return visit method, the push state of problem is paid a return visit according to target voice automatically for intelligent sound provided by the present embodiment Determine the need for push target pay a return visit phonetic problem, so as to push state be pushed state when, without pushing mesh again Poster sound pays a return visit problem, reduces volume of transmitted data and improves data-pushing efficiency.Target voice is being paid a return visit into problem or its is right After the problem of answering number and automatic speech outgoing call tables of data are pushed to telephony platform, telephony platform can be made to return based on the target voice Access topic or corresponding the returns visit voice data of its corresponding question number and the progress voice return visit of automatic speech outgoing call tables of data, Its process is not necessarily to manual intervention, can effectively improve the efficiency of voice return visit, reduces the cost of labor that voice is paid a return visit.
In one embodiment, as shown in fig. 6, carrying out text to outgoing call recording data using target voice static state decoding network Before the step of this translation, intelligent sound pays a return visit method automatically further include:
S601: training text data corresponding with voice return visit problem are obtained from training data corpus.
Wherein, training data corpus is the database for being exclusively used in storage training text data.Training text data are to deposit Store up the text data in training data corpus.Specifically, which can be is acquired on user by network The text data corresponding with voice return visit problem passed, and this article notebook data is stored in training data corpus.For example, In loan transaction, the training text data obtained from training data corpus may include returning for the corresponding voice of B02 Corresponding reply of access topic is " my amount of the loan is 500,000 ", pays a return visit the corresponding answer of problem for the corresponding voice of B03 and is " Time Of Release of my loan be on November 1st, 2017 ", then " my amount of the loan is 500,000 " and " when the granting of my loan Between be on November 1st, 2017 " be training text data.
S602: training text data are input to N-gram model and carry out model training, obtain target language model.
Wherein, N-gram is common a kind of language model in large vocabulary continuous speech recognition, using adjacent in context Collocation information between word, when needing the phonetic continuously without space to be converted into Chinese character string (i.e. sentence), can calculate has The sentence of maximum probability manually selects to realize the automatic conversion for arriving Chinese character without user, and it is one corresponding to avoid many Chinese characters Identical phonetic and lead to coincident code problem.N-gram is a kind of algorithm based on statistical language model.Its basic thought is will be literary Content inside this carries out the sliding window that size is n according to byte and operates, and forms the byte fragment sequence that length is n.Often One byte segment is known as gram, counts to the occurrence frequency of all gram, and according to the threshold value being previously set into Row filtering forms key gram list, that is, the vector characteristics space of this text, each gram in list is exactly one A feature vector dimension.
N-gram based on Markov assume: the appearance of n-th word only it is related to the word of front N-1, and with it is other any Word is all uncorrelated, and the probability of whole sentence is exactly the product of each word probability of occurrence.These probability can be by directly from all training N number of word is counted in text data while the number occurred obtains.That is P (T)=P (W1W2W3…Wn)=P (W1)P(W2|W1)P(W3| W1W2)…P(Wn|W1W2…Wn-1), wherein P (Wn|W1W2…Wn-1) refer to that n-th of participle appears in the language of n-1 participle composition Probability after sequence sequence.In N-gram model, usually using maximal possibility estimation (Maximum Likelihood Estimate) P (W is calculatedn|W1W2…Wn-1), i.e.,Wherein, C (Wn) it is the Word frequency of the n participle in all training text data, C (W1W2…Wn) it is (W1W2…Wn) sequence is in all training text data In word sequence frequency, C (W1W2…Wn-1) it is (W1W2…Wn-1) word sequence frequency of the sequence in all training text data.
In the present embodiment, by the training text data extracted in training data corpus be input to N-gram model into Row model training, so that the target language model obtained can be used for assessing whether the text identified meets user's reply Voice pays a return visit the model of the language use habit of problem.I.e. using N-gram model to the resulting target of training text data training Language model, the identification that can reply the outgoing call recording data that target voice pays a return visit problem to client are more acurrate.
S603: being based on training text data, acquires trained voice data corresponding with each training text data.
Specifically, it is previously provided with voice collecting tool in server, clicks " starting to record " in client in user After button, acquisition different user reads the voice when training text data, corresponding with each training text data to obtain Training voice data.It is to be appreciated that the training voice data is deposited after server often collects a training voice data It stores up in the database, the training sample as subsequent training objective acoustic model.
S604: training voice data is input to GMM-HMM model and carries out model training, obtains target acoustical model.
Wherein, target acoustical model is the acoustics obtained after being trained using training voice data to GMM-HMM model Model.Specifically, training voice data is input to GMM-HMM model and carries out model training by server, obtains target acoustical mould The process of type includes the following steps:
Firstly, carrying out feature extraction to training voice data, MFCC (Mel-frequency Cepstrum is obtained Coefficients, i.e. mel-frequency cepstrum coefficient) feature.Wherein, multidimensional characteristic vectors can be used in mel-frequency cepstrum coefficient The mode of (m dimension n column) is expressed, and it is a frame waveform, the corresponding state of several frame waveforms, every three combinations of states that m, which ties up 1 column vector, At a phoneme.
Then, GMM (Gaussian Mixed Model, gauss hybrid models) is trained using MFCC feature, is obtained Target GMM model is taken, process specifically includes: (1) initializing the parameter of GMM, which includes component number K, mixing Coefficient πk, mean μkWith covariance ∑k, point x={ x is formed by for all MFCC features1,x2,...,xN, GMM model For(2) using EM (Expectation Maximization Algorithm, Greatest hope) algorithm update GMM parameter, obtain target GMM.The EM algorithm includes E step and M Step.In E step, According to current mixed coefficint πk, mean μkWith covariance ∑k, calculate posterior probability γ (znk), whereinIn M step, according to the posterior probability γ (z being calculatednk), calculate new mix Collaboration number πk, mean μkWith covariance ∑k, target GMM model is obtained in parameter convergence, i.e.,Wherein,N is number a little.
Finally, target GMM model, which is input to HMM, carries out model training, target acoustical model is obtained, process is specifically wrapped It includes: (1) assuming to obey monokaryon gaussian probability distribution b in state observation sequencej(x)=p (xsj)=N (x；μj,∑j), initialization The parameter lambda of HMM, the parameter lambda include preceding to transition probability αij, back transition probability βt(sj), mean μjWith covariance ∑j, In, αijFor from state siIt is transferred to other states sjTransition probability, andβt(sj) it is that moment t is in state sj's Words, the probability of t moment future observation, i.e. βt(sj)=p (xt+1,xt+2,xT| s (t)=sj, λ), wherein αijFor from state siTurn Move on to other states sjTransition probability, bj(xt+1) it is to observe x under state it+1Probability, βt+1(sj) it is that t moment is in shape State sjIf, the probability observed after t+1.(2) the forward direction transition probability α of HMM is updated using EM algorithmij, mean μjAnd covariance ∑j, obtain target acoustical model.Process using parameter in EM algorithm update HMM is consistent with the process of parameter in GMM that updates, It will not repeat them here.
S605: being based on target language model and target acoustical model, constructs target voice static state decoding network.
Specifically, server is based on target language model, target acoustical model and pre-set pronunciation dictionary and acoustics Context, constructs target voice static state decoding network, and building process includes the following steps:
(1) by target language model, target acoustical model, pronunciation dictionary and acoustics context switch at WFST (Weighted Finite-state Transduce, weighted finite state converter) network, i.e., obtain language model respectively WFST (hereinafter referred to as G), pronunciation dictionary WFST (hereinafter referred to as L), acoustical context WFST (hereinafter referred to as C) and acoustics Model WFST (hereinafter referred to as H).It is a WFSA (acceptor receiver) in language model WFST, it can be with other three A WFST is operated, and an input symbol and the identical WFST of output symbol are regarded as, and is specifically defined word sequence appearance Probability.Pronunciation dictionary WFST, input symbol are monophone (phoneme), and output symbol is word.Pronunciation dictionary defines aligned phoneme sequence Represented word, according to the possible aligned phoneme sequence that across word triphone model generates, available corresponding word sequence.Acoustically Hereafter WFST, input symbol are triphone (three-tone), and output symbol is monophnoe (phoneme), the WFST net definitions Corresponding relationship from three-tone to phoneme, the three-tone sequence generated according to HMM model.Acoustic model WFST, input symbol are HMM transitions-ids (transition identifiers transformation identifier, be used to indicate corresponding feature vector), Output symbol is triphone (three-tone), defines HMM status switch corresponding to each three-tone.In speech recognition, lead to It crosses and state corresponding to each frame is carried out it is assumed that can scan on the status switch of HMM, to generate possible three Phone sequence.
(2) four WFST networks are merged and compression optimization, obtains target voice static state decoding network.
Specifically, first four WFST networks are merged using H °C ° L ° G °, obtain raw tone static state decoding network, Wherein, H is acoustic model WFST, and C is acoustical context WFST, and L is pronunciation dictionary WFST, and G is language model WFST, ° finger mould Type merges (Composition).Then, compression optimization is carried out to raw tone static state decoding network, it is static obtains target voice Decoding network.It since the committed memory of raw tone static state decoding network is very big, need to advanced optimize, so that the mesh after optimization Poster sound static state decoding network can have lesser volume.Specifically, using N=πε(min(det(H°det(C°det(L° G compression optimization))))) is carried out to raw tone static state decoding network, obtains target voice static state decoding network, so that it is formed Identification network it is smaller, wherein det (Determinization) be determinization algorithm, min (Minimization) be minimum Change algorithm, πεIt removes for idle running except (ε-Removal) algorithm.
In the present embodiment, acquired target voice static state decoding network allows probabilistic information in network using iterative calculation It transmits and updates between node, to carry out tone decoding, and since static decoding network has all been unfolded search space, because This, does not need to construct search space copy according to forerunner's word of decoding paths, not need in suffix node according to historical information yet Query language model, so that decoding speed is very fast when its subsequent progress speech recognition.
In one embodiment, according to the corresponding input/output relation of four WFST it is found that target voice is quiet in step S205 State decoding network carries out text translation to outgoing call recording data, obtains outgoing call text data, specifically comprises the following steps: that (1) is adopted Outgoing call recording data is handled with acoustic model WFST, obtains HMM status switch, every HMM status switch is one The status switch of triphone (three-tone).HMM (Hidden Markov Model, hidden Markov model) is preparatory training Good state network, each frame voice data that will acquire belong in each shape probability of state input HMM, can be from the state net Most coupling path between frame and state is found in network, will the corresponding state of most coupling path as dbjective state, from state network Most coupling path between searching state and phoneme, according to this, most coupling path determines aligned phoneme sequence.Wherein, route searching in HMM Algorithm is a kind of algorithm of Dynamic Programming beta pruning, referred to as Viterbi (Viterbi) algorithm, for finding global optimum path, And then realize that by MFCC Feature Conversion be HMM status switch.(2) at using acoustical context WFST to HMM status switch Reason, obtains corresponding aligned phoneme sequence.Acoustical context WFST defines three-tone to the corresponding relationship of phoneme, therefore, can be used Acoustical context WFST handles HMM status switch, obtains corresponding aligned phoneme sequence.(3) WFST pairs of pronunciation dictionary is used Aligned phoneme sequence is handled, and corresponding word sequence is obtained.(4) word sequence is handled using language model WFST, is obtained outer Exhale text data.
It is to be appreciated that due to acoustic model WFST, acoustical context WFST, pronunciation dictionary WFST and language model WFST It is four concatenated subsystems in trained target voice static state decoding network, the output of each subsystem is next height The input of system, so that being merged to four WFST, determinization, minimum and idle running remove the static decoding obtained except after Outgoing call recording data directly can be input to acoustic model WFST by network, successively pass through acoustical context WFST, pronunciation dictionary WFST and language model WFST processing, can obtain corresponding outgoing call text data, in decoding process due to will search it is empty Between be all unfolded, can be used Viterbi (Viterbi) algorithm quick obtaining it is optimal decoding as a result, making its decoding speed fast.
In one embodiment, as shown in step S202, the target answer data in issue database is in issue database A pre-stored answer data, specifically include question number, at least two option names and with each option names phase Corresponding option number.Wherein, option number is the number of its corresponding option names for unique identification.For example, for asking The voice that topic number is QT01 pays a return visit problem, the option names in answer data include satisfied, dissatisfied, general and client not Cooperate this four options, respectively corresponds this four option numbers of QP01, QP02, QP03 and QP04.It is to be appreciated that target answer Each option names in data are that a kind of answer of problem is paid a return visit for target voice.Optional understanding ground, for each choosing Item title, can configure corresponding option keyword, be mentioned to corresponding choosing to reply when target voice pays a return visit problem in user Item keyword, that is, can determine its selected option names.As shown in fig. 7, corresponding outer to problem is paid a return visit with target voice It exhales text data and target answer data to be analyzed, obtains automatic speech outgoing call result, comprising:
S701: keyword extraction is carried out to outgoing call text data corresponding with target voice return visit problem, obtains target Keyword.
Specifically, server first can carry out word segmentation processing to outgoing call text data after obtaining outgoing call text data, with Obtain multiple participles；Again using going deactivated word algorithm that the corresponding all participles of text data is externally exhaled to carry out identifying processings, to go Except stop words, it regard remaining participle as target keyword, this mode obtains the process of target keyword simply and is easy to real It is existing.
S702: by the option names in target keyword and target answer data corresponding with target voice return visit problem Matching treatment is carried out, the targets option title of successful match is obtained.
Specifically, server is after obtaining target keyword from outgoing call text data, using string matching algorithm pair The corresponding choosing of each option names in target keyword target answer data corresponding with target voice return visit problem Item keyword carries out matching treatment, obtains the option keyword of successful match, and the option keyword of successful match is corresponding Option names are determined as targets option title.The string matching algorithm includes but is not limited to BF (Brute Force, violence inspection Rope), RK (Robin-Karp, Hash retrieval), KMP (The Knuth-Morris-Pratt Algorithm, Alexandre Desplat) etc. calculate Method.
S703: targets option title is mapped in default result analysis template in corresponding optional domain, automatic language is obtained Sound outgoing call result.
Default result analysis template be stored in advance in the server for carrying out automatic speech outgoing call interpretation of result Template.In the present embodiment, server, first will be in the automatic speech outgoing call tables of data after obtaining automatic speech outgoing call tables of data Customer data be mapped to default result analysis template on, the default result analysis template in also set up each target voice return Corresponding question number field and optional domain corresponding with the question number field are inscribed in access, which is for filling The region of value corresponding with question number field.Correspondingly, server is in the outgoing call recording data for receiving telephony platform feedback And be converted into after outgoing call text data based on the outgoing call recording data, each target language is got using step S701-S702 Sound pays a return visit the corresponding targets option title of problem, and targets option title mapping (filling) is asked to target voice return visit In the corresponding optional domain of the problem of topic number field, to obtain automatic speech outgoing call as a result, making the automatic speech outgoing call knot Fruit can be intuitive to see that the corresponding client of customer data is to the various feedbacks for the business handled in automatic speech outgoing call tables of data Opinion facilitates carry out business improvement.
Further, it is additionally provided in default result analysis template and summarizes field and to summarize field corresponding with this Optional domain, each to summarize the corresponding numerical computational formulas of field, which can refer to question number field pair The quantity of each targets option title in the optional domain answered, be also possible to quantity based on each option names calculate satisfaction or The calculation formula of other indexs of person.Targets option title is mapped in server corresponding with corresponding question number field After optional domain, the data in optional domain are automatically analyzed using the numerical computational formulas, to obtain corresponding total amount According to, then this is summarized into data be mapped to and summarize in the corresponding optional domain of field with this, to obtain last automatic speech outgoing call As a result.The automatic speech outgoing call result can intuitively show that client to the various feedback opinions for the business handled, and is converged Bulk analysis facilitates carry out business improvement.
Intelligent sound provided by the present embodiment in return visit method, is mentioned automatically by carrying out keyword to outgoing call text data Take, can the corresponding target keyword of quick obtaining, and the target keyword and target voice are paid a return visit into the corresponding mesh of problem The option names marked in answer data carry out matching treatment, to determine the targets option title of successful match, to return to user The outgoing call text data that the multiple target voice pays a return visit problem is standardized, and to carry out data analysis, improves data point Analyse efficiency.Then, which is mapped in default result analysis template, automatic speech outgoing call knot can be directly acquired Fruit improves analysis efficiency and reduces cost of labor so that the analytic process of automatic speech outgoing call result is not necessarily to manual intervention.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit It is fixed.
In one embodiment, a kind of automatic playback appliances of intelligent sound are provided, the automatic playback appliances of the intelligent sound with it is upper It states intelligent sound in embodiment and pays a return visit method one-to-one correspondence automatically.As shown in figure 8, the automatic playback appliances of the intelligent sound include language Sound callback access request obtains module 801, target pays a return visit problem and obtains module 802, outgoing call tables of data acquisition module 803, data-pushing Processing module 804, outgoing call text data obtain module 805 and voice outgoing call result obtains module 806.Each functional module is specifically It is bright as follows:
Voice callback access request obtains module 801, and for obtaining voice callback access request, voice callback access request includes target channel Mark；
Target return visit problem obtains module 802, for inquiring issue database, acquisition and target based on target channel mark The corresponding target voice of channel mark pays a return visit problem and target answer data corresponding with target voice return visit problem；
Outgoing call tables of data obtains module 803, for being based on target channel mark inquiring customer information database, acquisition and mesh Mark the corresponding automatic speech outgoing call tables of data of channel mark；
Data-pushing processing module 804 is pushed to for automatic speech outgoing call tables of data and target voice to be paid a return visit problem Telephony platform, so that telephony platform is based on return visit voice data corresponding with target voice return visit problem and automatic speech outgoing call number Voice return visit is carried out according to table, obtains the outgoing call recording data corresponding with target voice return visit problem that telephony platform returns；
Outgoing call text data obtain module 805, for using target voice static state decoding network to outgoing call recording data into It composes a piece of writing this translation, obtains outgoing call text data corresponding with target voice return visit problem；
Voice outgoing call result obtains module 806, for outgoing call text data corresponding with target voice return visit problem It is analyzed with target answer data, obtains automatic speech outgoing call result.
Preferably, before voice callback access request obtains module 801, the automatic playback appliances of intelligent sound further include outgoing call name Forms data acquiring unit, outgoing call list data acquiring unit, first judge that processing unit and second judges processing unit.
Outgoing call list data acquiring unit, for obtaining outgoing call list data, outgoing call list data includes that at least one is new Increase customer data, newly-increased customer data includes customer ID.
Existing customer data judging unit judges customer information for being based on customer ID inquiring customer information database Database whether there is existing customer data corresponding with customer ID.
First judges processing unit, if existing client corresponding with customer ID is not present for customer information database Newly-increased customer data is then stored in customer information database by data, is configured the newly-increased corresponding task status of customer data and is New task state.
Second judges processing unit, if for there is existing client corresponding with customer ID in customer information database Data then update existing customer data using newly-increased customer data, and retain the existing corresponding task status of customer data.
Preferably, it includes voice outgoing call task acquiring unit, the acquisition of effective customer data that outgoing call tables of data, which obtains module 803, Unit, outgoing call issue mode acquiring unit, automatic outer call processing unit and artificial outgoing call processing unit.
Voice outgoing call task acquiring unit, for obtaining voice outgoing call task, voice outgoing call task includes target channel mark Know and goal task state.
Effective customer data acquiring unit, for being based on target channel mark and goal task status inquiry customer information number According to library, effective customer data is obtained.
Outgoing call issues mode acquiring unit, for being based on target channel mark and goal task state, determines optional outgoing call Mode is issued, and mode is issued according to optional outgoing call and determines that target outgoing call issues mode.
Automatic outer call processing unit, for obtaining target when it is that automatic outer call issues mode that target outgoing call, which issues mode, Date and destination name odd number amount are pushed, if the target push date meets default push rule, is chosen and destination name odd number amount one The effective customer data caused, as automatic speech outgoing call tables of data corresponding with target channel mark.
Artificial outgoing call processing unit, for obtaining target when it is that artificial outgoing call issues mode that target outgoing call, which issues mode, List quantity, selection and the consistent effective customer data of destination name odd number amount, as corresponding with target channel mark automatic Voice outgoing call tables of data.
Preferably, data-pushing processing module 804 includes push state acquiring unit, the first push processing unit and second Push processing unit.
Push state acquiring unit obtains target voice and pays a return visit for inquiring issue database based on target channel mark The corresponding push state of problem.
First push processing unit, if being to have pushed state for pushing state, by automatic speech outgoing call tables of data and Target voice pays a return visit the corresponding question number of problem and is pushed to telephony platform, so that telephony platform is based on question number corresponding time It visits voice data and automatic speech outgoing call tables of data carries out voice return visit.
Second push processing unit, if being not push state for pushing state, by automatic speech outgoing call tables of data and Target voice pays a return visit problem and is pushed to telephony platform, so that telephony platform is based on, target voice return visit problem acquisition is corresponding to be paid a return visit Voice data, and voice return visit is carried out based on return visit voice data and automatic speech outgoing call tables of data.
Preferably, before outgoing call text data obtains module 805, the automatic playback appliances of intelligent sound further include training text Notebook data acquiring unit, target language model acquiring unit, training voice data acquiring unit, target acoustical model acquiring unit With static decoding network acquiring unit.
Training text data capture unit, it is corresponding with voice return visit problem for being obtained from training data corpus Training text data.
Target language model acquiring unit carries out model training for training text data to be input to N-gram model, Obtain target language model.
Training voice data acquiring unit acquires opposite with each training text data for being based on training text data The training voice data answered.
Target acoustical model acquiring unit carries out model training for voice data will to be trained to be input to GMM-HMM model, Obtain target acoustical model.
Static decoding network acquiring unit constructs target voice for being based on target language model and target acoustical model Static decoding network.
Preferably, target answer data includes at least two option names.
Voice outgoing call result obtain module 806 include target keyword acquiring unit, target name acquiring unit to be recommended and Voice outgoing call result acquiring unit.
Target keyword acquiring unit, for being closed to outgoing call text data corresponding with target voice return visit problem Keyword extracts, and obtains target keyword.
Target name acquiring unit to be recommended is used for target keyword and target corresponding with target voice return visit problem Option names in answer data carry out matching treatment, obtain the targets option title of successful match.
Voice outgoing call result acquiring unit, it is corresponding in default result analysis template for targets option title to be mapped to In optional domain, automatic speech outgoing call result is obtained.
Specific restriction about the automatic playback appliances of intelligent sound may refer to pay a return visit automatically above for intelligent sound The restriction of method, details are not described herein.Modules in the above-mentioned automatic playback appliances of intelligent sound can be fully or partially through Software, hardware and combinations thereof are realized.Above-mentioned each module can be embedded in the form of hardware or independently of the place in computer equipment It manages in device, can also be stored in a software form in the memory in computer equipment, in order to which processor calls execution or more The corresponding operation of modules.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junction Composition can be as shown in Figure 9.The computer equipment include by system bus connect processor, memory, network interface and Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating The database of machine equipment executes intelligent sound in above-described embodiment for server and pays a return visit method use or shape in the process automatically At data.The network interface of the computer equipment is used to communicate with external terminal by network connection.The computer program To realize that a kind of intelligent sound pays a return visit method automatically when being executed by processor.
In one embodiment, a kind of computer equipment is provided, including memory, processor and storage are on a memory And the computer program that can be run on a processor, processor realize user account number in above-described embodiment when executing computer program The step of unlocking method, such as step S201-S206 or Fig. 3 shown in Fig. 2 is to step shown in fig. 7.Alternatively, processing Device realizes the function of each module/unit in this embodiment of user account number tripper, such as Fig. 8 when executing computer program Shown in voice callback access request obtain module 801 to voice outgoing call result obtain module 806 function, to avoid repeating, here It repeats no more.
In one embodiment, a computer readable storage medium is provided, meter is stored on the computer readable storage medium The step of calculation machine program, which realizes user account number unlocking method in above-described embodiment when being executed by processor, example Step S201-S206 or Fig. 3 as shown in Figure 2 is to step shown in fig. 7, and to avoid repeating, which is not described herein again.Or Person, the computer program realize each module in above-mentioned this embodiment of user account number tripper/mono- when being executed by processor The function of member, such as voice callback access request shown in Fig. 8 obtain the function that module 801 obtains module 806 to voice outgoing call result, To avoid repeating, which is not described herein again.
Those of ordinary skill in the art will appreciate that realizing all or part of the process of above-described embodiment method, can pass through Computer program is completed to instruct relevant hardware, computer program can be stored in a non-volatile computer and can be read and deposit In storage media, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Provided herein To any reference of memory, storage, database or other media used in each embodiment, may each comprise non-volatile And/or volatile memory.Nonvolatile memory may include that read-only memory (ROM), programming ROM (PROM), electricity can be compiled Journey ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) directly RAM (RDRAM), straight Connect memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM).
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations；Although referring to aforementioned reality Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features；And these are modified Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all It is included within protection scope of the present invention.

Claims (10)

1. a kind of intelligent sound pays a return visit method automatically characterized by comprising
Voice callback access request is obtained, the voice callback access request includes target channel mark；
Issue database is inquired based on the target channel mark, obtains target voice corresponding with the target channel mark Return visit problem and target answer data corresponding with target voice return visit problem；
Based on the target channel mark inquiring customer information database, obtain corresponding with the target channel mark automatic Voice outgoing call tables of data；
The automatic speech outgoing call tables of data and the target voice are paid a return visit into problem and are pushed to telephony platform, so that the phone Platform is based on paying a return visit the corresponding return visit voice data of problem with the target voice and the automatic speech outgoing call tables of data carries out Voice is paid a return visit, and the outgoing call recording data corresponding with target voice return visit problem that the telephony platform returns is obtained；
Text translation is carried out to the outgoing call recording data using target voice static state decoding network, is obtained and the target voice The corresponding outgoing call text data of return visit problem；
It analyzes with the target voice return visit corresponding outgoing call text data of problem and the target answer data, obtains Take automatic speech outgoing call result.
2. intelligent sound as described in claim 1 pays a return visit method automatically, which is characterized in that in the acquisition voice callback access request The step of before, the intelligent sound pays a return visit method automatically further include:
Outgoing call list data is obtained, the outgoing call list data includes at least one newly-increased customer data, newly-increased client's number According to including customer ID；
The customer information database is inquired based on the customer ID, judges that the customer information database whether there is and institute State the corresponding existing customer data of customer ID；
It, will be described newly-increased if existing customer data corresponding with the customer ID is not present in the customer information database Customer data is stored in the customer information database, and configuring the corresponding task status of the newly-increased customer data is new task State；
If there is existing customer data corresponding with the customer ID in the customer information database, using described new Increase customer data and update the existing customer data, and retains the corresponding task status of the existing customer data.
3. intelligent sound as described in claim 1 pays a return visit method automatically, which is characterized in that described to be based on the target channel mark Know inquiring customer information database, obtain automatic speech outgoing call tables of data corresponding with the target channel mark, comprising:
Voice outgoing call task is obtained, the voice outgoing call task includes target channel mark and goal task state；
Based on the target channel mark and the goal task status inquiry customer information database, effective client's number is obtained According to；
Based on the target channel mark and the goal task state, determine that optional outgoing call issues mode, and according to it is described can It selects outgoing call to issue mode and determines that target outgoing call issues mode；
When it is that automatic outer call issues mode that target outgoing call, which issues mode, target push date and destination name odd number amount are obtained, if The target push date meets default push rule, then selection and the consistent effective customer data of the destination name odd number amount, As automatic speech outgoing call tables of data corresponding with the target channel mark；
When it is that artificial outgoing call issues mode that target outgoing call, which issues mode, destination name odd number amount, selection and the destination name are obtained Odd number measures consistent effective customer data, as automatic speech outgoing call tables of data corresponding with the target channel mark.
4. intelligent sound as described in claim 1 pays a return visit method automatically, which is characterized in that described by the automatic speech outgoing call Tables of data and the target voice pay a return visit problem and are pushed to telephony platform, so that the telephony platform is based on and the target voice The corresponding return visit voice data of return visit problem and the automatic speech outgoing call tables of data carry out voice return visit, comprising:
Issue database is inquired based on the target channel mark, the target voice is obtained and pays a return visit the corresponding push shape of problem State；
If the push state is to have pushed state, the automatic speech outgoing call tables of data and target voice return visit are asked It inscribes corresponding question number and is pushed to telephony platform, so that the telephony platform is based on described problem and numbers corresponding return visit voice Data and the automatic speech outgoing call tables of data carry out voice return visit；
If the push state is not push state, the automatic speech outgoing call tables of data and target voice return visit are asked Topic is pushed to telephony platform, so that the telephony platform is based on, target voice return visit problem acquisition is corresponding to pay a return visit voice number According to, and voice return visit is carried out based on the return visit voice data and the automatic speech outgoing call tables of data.
5. intelligent sound as described in claim 1 pays a return visit method automatically, which is characterized in that described static using target voice Before the step of decoding network carries out text translation to the outgoing call recording data, the intelligent sound is paid a return visit method automatically and is also wrapped It includes:
Training text data corresponding with voice return visit problem are obtained from training data corpus；
The training text data are input to N-gram model and carry out model training, obtain target language model；
Based on the training text data, acquisition trained voice data corresponding with each training text data；
The trained voice data is input to GMM-HMM model and carries out model training, obtains target acoustical model；
Based on the target language model and the target acoustical model, target voice static state decoding network is constructed.
6. intelligent sound as described in claim 1 pays a return visit method automatically, which is characterized in that the target answer data includes extremely Few two option names；
It is described to divide with the target voice return visit corresponding outgoing call text data of problem and the target answer data Analysis obtains automatic speech outgoing call result, comprising:
Keyword extraction is carried out to outgoing call text data corresponding with target voice return visit problem, obtains target critical Word；
By the option names in the target keyword and target answer data corresponding with target voice return visit problem Matching treatment is carried out, the targets option title of successful match is obtained；
The targets option title is mapped in default result analysis template in corresponding optional domain, automatic speech outgoing call is obtained As a result.
7. a kind of automatic playback appliances of intelligent sound characterized by comprising
Voice callback access request obtains module, and for obtaining voice callback access request, the voice callback access request includes target channel mark Know；
Target pays a return visit problem and obtains module, for inquiring issue database based on the target channel mark, obtains and the mesh It marks the corresponding target voice of channel mark and pays a return visit problem and target answer number corresponding with target voice return visit problem According to；
Outgoing call tables of data obtain module, for be based on the target channel mark inquiring customer information database, obtain with it is described The corresponding automatic speech outgoing call tables of data of target channel mark；
Data-pushing processing module is pushed to for the automatic speech outgoing call tables of data and the target voice to be paid a return visit problem Telephony platform so that the telephony platform be based on the target voice pay a return visit corresponding the returns visit voice data of problem and it is described oneself Dynamic voice outgoing call tables of data carries out voice return visit, obtains the opposite with target voice return visit problem of the telephony platform return The outgoing call recording data answered；
Outgoing call text data obtains module, for carrying out text to the outgoing call recording data using target voice static state decoding network This translation obtains outgoing call text data corresponding with target voice return visit problem；
Voice outgoing call result obtains module, for outgoing call text data corresponding with target voice return visit problem and institute It states target answer data to be analyzed, obtains automatic speech outgoing call result.
8. the automatic playback appliances of intelligent sound as claimed in claim 7, which is characterized in that the target answer data includes extremely Few two option names；The voice outgoing call result obtains module
Target keyword acquiring unit, for being closed to outgoing call text data corresponding with target voice return visit problem Keyword extracts, and obtains target keyword；
Target name acquiring unit to be recommended, it is corresponding for paying a return visit problem by the target keyword and with the target voice Option names in target answer data carry out matching treatment, obtain the targets option title of successful match；
Voice outgoing call result acquiring unit, it is corresponding in default result analysis template for the targets option title to be mapped to In optional domain, automatic speech outgoing call result is obtained.
9. a kind of computer equipment, including memory, processor and storage are in the memory and can be in the processor The computer program of upper operation, which is characterized in that the processor realized when executing the computer program as claim 1 to The step of any one of 6 intelligent sounds pay a return visit method automatically.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists In realizing the intelligent sound side of return visit automatically as described in any one of claim 1 to 6 when the computer program is executed by processor The step of method.
CN201811010990.1A 2018-08-31 2018-08-31 Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically Pending CN109274845A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811010990.1A CN109274845A (en) 2018-08-31 2018-08-31 Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811010990.1A CN109274845A (en) 2018-08-31 2018-08-31 Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically

Publications (1)

Publication Number Publication Date
CN109274845A true CN109274845A (en) 2019-01-25

Family

ID=65155108

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811010990.1A Pending CN109274845A (en) 2018-08-31 2018-08-31 Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically

Country Status (1)

Country Link
CN (1) CN109274845A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084616A (en) * 2019-03-15 2019-08-02 深圳壹账通智能科技有限公司 Intelligence pays a return visit method, apparatus, computer installation and storage medium
CN110765776A (en) * 2019-10-11 2020-02-07 阳光财产保险股份有限公司 Method and device for generating return visit labeling sample data
CN111063359A (en) * 2019-12-24 2020-04-24 太平金融科技服务（上海）有限公司 Telephone return visit validity judging method, device, computer equipment and medium
CN111246031A (en) * 2020-02-27 2020-06-05 大连即时智能科技有限公司 Man-machine cooperative telephone customer service method and system
CN112101849A (en) * 2020-08-11 2020-12-18 大亚湾核电运营管理有限责任公司 Automatic nuclear power station data verification method, device, equipment and medium
CN113313386A (en) * 2021-05-31 2021-08-27 浙江惠瀜网络科技有限公司 Intelligent voice investigation system and method for automobile financial risk
CN113395395A (en) * 2021-05-31 2021-09-14 浙江惠瀜网络科技有限公司 AI-based intelligent voice return visit method
WO2022126964A1 (en) * 2020-12-15 2022-06-23 平安科技（深圳）有限公司 Service data verification method and apparatus, device and storage medium
CN115878764A (en) * 2022-03-07 2023-03-31 北京中关村科金技术有限公司 Voice return access questionnaire investigation method and system, computing equipment and storage medium

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040240655A1 (en) * 2003-05-30 2004-12-02 Swick Corvin F. Technique for assisting a user with information and customer services at a call center
CN102572134A (en) * 2010-12-30 2012-07-11 上海博泰悦臻电子设备制造有限公司 Work order processing method, work order building system and work order processing system
CN103065630A (en) * 2012-12-28 2013-04-24 安徽科大讯飞信息科技股份有限公司 User personalized information voice recognition method and user personalized information voice recognition system
CN103680500A (en) * 2012-08-29 2014-03-26 北京百度网讯科技有限公司 Speech recognition method and device
CN105376433A (en) * 2015-09-24 2016-03-02 深圳市车音网科技有限公司 Voice automatic revisit device, system and method
CN105609107A (en) * 2015-12-23 2016-05-25 北京奇虎科技有限公司 Text processing method and device based on voice identification
CN105976812A (en) * 2016-04-28 2016-09-28 腾讯科技（深圳）有限公司 Voice identification method and equipment thereof
CN106254696A (en) * 2016-08-02 2016-12-21 北京京东尚科信息技术有限公司 Outgoing call result determines method, Apparatus and system
CN106303102A (en) * 2015-06-25 2017-01-04 阿里巴巴集团控股有限公司 Automatization's calling-out method, Apparatus and system
CN106412349A (en) * 2015-07-16 2017-02-15 中兴通讯股份有限公司 Method and system of realizing outbound call system questionnaire flows
CN106683677A (en) * 2015-11-06 2017-05-17 阿里巴巴集团控股有限公司 Method and device for recognizing voice
CN106816148A (en) * 2015-11-30 2017-06-09 三星电子株式会社 Speech recognition apparatus and method
CN107026949A (en) * 2017-02-20 2017-08-08 平安科技（深圳）有限公司 Pay a return visit the treating method and apparatus of voice messaging
CN107644638A (en) * 2017-10-17 2018-01-30 北京智能管家科技有限公司 Audio recognition method, device, terminal and computer-readable recording medium
CN108305634A (en) * 2018-01-09 2018-07-20 深圳市腾讯计算机系统有限公司 Coding/decoding method, decoder and storage medium

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040240655A1 (en) * 2003-05-30 2004-12-02 Swick Corvin F. Technique for assisting a user with information and customer services at a call center
CN102572134A (en) * 2010-12-30 2012-07-11 上海博泰悦臻电子设备制造有限公司 Work order processing method, work order building system and work order processing system
CN103680500A (en) * 2012-08-29 2014-03-26 北京百度网讯科技有限公司 Speech recognition method and device
CN103065630A (en) * 2012-12-28 2013-04-24 安徽科大讯飞信息科技股份有限公司 User personalized information voice recognition method and user personalized information voice recognition system
CN106303102A (en) * 2015-06-25 2017-01-04 阿里巴巴集团控股有限公司 Automatization's calling-out method, Apparatus and system
CN106412349A (en) * 2015-07-16 2017-02-15 中兴通讯股份有限公司 Method and system of realizing outbound call system questionnaire flows
CN105376433A (en) * 2015-09-24 2016-03-02 深圳市车音网科技有限公司 Voice automatic revisit device, system and method
CN106683677A (en) * 2015-11-06 2017-05-17 阿里巴巴集团控股有限公司 Method and device for recognizing voice
CN106816148A (en) * 2015-11-30 2017-06-09 三星电子株式会社 Speech recognition apparatus and method
CN105609107A (en) * 2015-12-23 2016-05-25 北京奇虎科技有限公司 Text processing method and device based on voice identification
CN105976812A (en) * 2016-04-28 2016-09-28 腾讯科技（深圳）有限公司 Voice identification method and equipment thereof
CN106254696A (en) * 2016-08-02 2016-12-21 北京京东尚科信息技术有限公司 Outgoing call result determines method, Apparatus and system
CN107026949A (en) * 2017-02-20 2017-08-08 平安科技（深圳）有限公司 Pay a return visit the treating method and apparatus of voice messaging
CN107644638A (en) * 2017-10-17 2018-01-30 北京智能管家科技有限公司 Audio recognition method, device, terminal and computer-readable recording medium
CN108305634A (en) * 2018-01-09 2018-07-20 深圳市腾讯计算机系统有限公司 Coding/decoding method, decoder and storage medium

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110084616A (en) * 2019-03-15 2019-08-02 深圳壹账通智能科技有限公司 Intelligence pays a return visit method, apparatus, computer installation and storage medium
CN110765776A (en) * 2019-10-11 2020-02-07 阳光财产保险股份有限公司 Method and device for generating return visit labeling sample data
CN111063359A (en) * 2019-12-24 2020-04-24 太平金融科技服务（上海）有限公司 Telephone return visit validity judging method, device, computer equipment and medium
CN111063359B (en) * 2019-12-24 2022-03-18 太平金融科技服务（上海）有限公司 Telephone return visit validity judging method, device, computer equipment and medium
CN111246031A (en) * 2020-02-27 2020-06-05 大连即时智能科技有限公司 Man-machine cooperative telephone customer service method and system
CN112101849A (en) * 2020-08-11 2020-12-18 大亚湾核电运营管理有限责任公司 Automatic nuclear power station data verification method, device, equipment and medium
CN112101849B (en) * 2020-08-11 2023-12-01 大亚湾核电运营管理有限责任公司 Nuclear power station data automatic verification method, device, equipment and medium
WO2022126964A1 (en) * 2020-12-15 2022-06-23 平安科技（深圳）有限公司 Service data verification method and apparatus, device and storage medium
CN113313386A (en) * 2021-05-31 2021-08-27 浙江惠瀜网络科技有限公司 Intelligent voice investigation system and method for automobile financial risk
CN113395395A (en) * 2021-05-31 2021-09-14 浙江惠瀜网络科技有限公司 AI-based intelligent voice return visit method
CN115878764A (en) * 2022-03-07 2023-03-31 北京中关村科金技术有限公司 Voice return access questionnaire investigation method and system, computing equipment and storage medium
CN115878764B (en) * 2022-03-07 2023-08-11 北京中关村科金技术有限公司 Voice return visit volume investigation method and system, computing device and storage medium

Similar Documents

Publication Publication Date Title
CN109274845A (en) Intelligent sound pays a return visit method, apparatus, computer equipment and storage medium automatically
CN109151218A (en) Call voice quality detecting method, device, computer equipment and storage medium
US10643604B2 (en) Language model customization in speech recognition for speech analytics
US10157609B2 (en) Local and remote aggregation of feedback data for speech recognition
CN110675288B (en) Intelligent auxiliary judgment method, device, computer equipment and storage medium
US8219404B2 (en) Method and apparatus for recognizing a speaker in lawful interception systems
US7966171B2 (en) System and method for increasing accuracy of searches based on communities of interest
US9361891B1 (en) Method for converting speech to text, performing natural language processing on the text output, extracting data values and matching to an electronic ticket form
US8880403B2 (en) Methods and systems for obtaining language models for transcribing communications
US10498888B1 (en) Automatic call classification using machine learning
CN110704571B (en) Court trial auxiliary processing method, trial auxiliary processing device, equipment and medium
US20110004473A1 (en) Apparatus and method for enhanced speech recognition
CN108989341A (en) The autonomous register method of voice, device, computer equipment and storage medium
CN111182162B (en) Telephone quality inspection method, device, equipment and storage medium based on artificial intelligence
US20100106499A1 (en) Methods and apparatus for language identification
CN108989349A (en) User account number unlocking method, device, computer equipment and storage medium
CN110347810A (en) Method, apparatus, computer equipment and storage medium are answered in dialog mode retrieval
Lee et al. On natural language call routing
CN109346082A (en) Sales order acquisition methods, device, equipment and medium based on speech recognition
CN114330366A (en) Event extraction method and related device, electronic equipment and storage medium
Kamm et al. Speech recognition issues for directory assistance applications
KR100803900B1 (en) Speech recognition ars service method, and speech recognition ars service system
CN116680360A (en) Smart customer service method and system
WO2019132690A1 (en) Method and device for building voice model of target speaker
CN110209804B (en) Target corpus determining method and device, storage medium and electronic device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20190125

RJ01 Rejection of invention patent application after publication