CN103686723A

CN103686723A - System and method for processing complex SMS (short message service) message of mobile customer hotline SMS messaging service hall

Info

Publication number: CN103686723A
Application number: CN201210347056.5A
Authority: CN
Inventors: 王卫民; 王石; 符建辉
Original assignee: KNOWOLOGY INTELLIGENT TECHNOLOGY Co Ltd
Current assignee: KNOWOLOGY INTELLIGENT TECHNOLOGY Co Ltd
Priority date: 2012-09-19
Filing date: 2012-09-19
Publication date: 2014-03-26

Abstract

The invention relates to a system and method for processing complex an SMS message of a mobile customer hotline SMS messaging service hall. The system comprises a login module for entering an operator reply module and a knowledge base maintenance module after passing validation of login information, an SMS message access module for receiving SMS messages incapable of being understood by the SMS messaging service hall, a natural language understanding module for understanding the content of an SMS message, a knowledge base module for managing mobile knowledge, an automatic replay module for automatically replaying SMS messages with high recognition rate, an operator replay module for manually reviewing and replying natural language understanding results by an operator, with the replaying content coming from the knowledge base, an answer knowledge search module for acquiring answer knowledge according to the SMS information and the natural language understanding results, a knowledge base searching module for inputting a natural language problem and searching the desired answer knowledge, a knowledge template instantiation module for obtaining a concrete knowledge answer, and a trigger operation module for being triggered to complete execution when the SMS content contains the execution requirement of a business action.

Description

Treatment system and the method thereof of the complicated note in the mobile client hot line short message service Room

Technical field

The present invention relates to extensive knowledge processing, mobile client service, natural language processing and automatic management field, particularly relate to the not treatable complicated short message process system in a kind of mobile client hot line short message service Room and method thereof.

Background technology

The variety of issue and the difficulty that for solving mobile client, in using mobile phone process, run into, China Mobile has set up customer service hotline throughout the country.Client can send at any time note to client's Consulting Hotlines and handle miscellaneous service, grasps the telephone expenses situation of oneself etc.SMS business hall (being called for short the short Room) is exactly the responsible existing system that sends to the note of client's hot line to process to client.It can process the short message content of the format of client's transmission, helps user to complete and wants the business of carrying out to move, and while making client want to handle or quit the subscription of a certain business, only need send corresponding contents to client's hot line.

Yet there are following three distinct issues in existing short message service Room system: one, the natural language processing in the short message service Room is limited in one's ability, and it can only understand consulting word or the instruction of some set form; Two, user sends to a lot of notes of mobile client hot line, is all to inquire for mobile communication relevant knowledge, or to mobile operator present one's view suggestion and complaint etc., and the existing short Room of these notes system can not normal process.Three, short Room system can only be processed the consulting of the note channel sending over by mobile customer service hot line, the note of other electronic channel, and short message service Room system does not have disposal ability.

Simultaneously, the IKBS that our company has developed, there is very strong natural language processing ability, Knowledge Management Capability, not only can process the consulting of note channel, also can treatments B BS, WAP etc. the consulting of other various electronic channels, but Intelligence repository lacks the ability of execution business action.Cause client to send to the demand of the execution business action of IKBS to be not being met.

Along with the development of mobile service, the contradiction of the expanding day of mobile client object and demand and the short message service Room and Intelligence repository scarce capacity is more and more obvious.Be badly in need of a kind of new system, it can process the not treatable complicated note in the short Room, indirectly strengthens service ability and the natural language processing ability in the short message service Room, meanwhile, can make again Intelligence repository have the ability of execution business action.

Summary of the invention

For above problem, the invention provides a kind of format consulting that not only can process access, also be appreciated that and process complicated unformatted reference content, not only can process the not treatable consultings in the short message service Room such as knowledge interrogators class, also can processing execution action class seek advice from, not only can process the consulting of note channel class, also can process the not treatable complicated short message process system of a kind of mobile client hot line SMS business hall and the method thereof of mobile other channel consultings.Object is to meet the growing demand of mobile client object, improves mobile hot line in the customer satisfaction of interior various electronic channels and the efficiency of mobile customer service.

In order to overcome the above problems, the invention provides the treatment system of the complicated note in a kind of mobile client hot line short message service Room, comprise login module, note access module, natural language processing module, automatically reply module, operator replys module, trigger action module, answer search module, knowledge base searching module, answer template instances module, knowledge base maintenance module;

Login module, for verifying log-on message, operator by authentication after, enter operator and reply module, system user by authentication after, enter knowledge base maintenance module, according to the role's of system user difference, be divided into following several situation:

1. knowledge base editorial staff logins by individual job number and password, by entering movable service knowledge base after authentication;

2. mobile contact staff logins by individual job number and password, by entering movable service knowledge base after authentication;

Note access module is the interface module of the treatment system of the complicated note in the short message service Room and the mobile client hot line short message service Room; First be that client sends note to client's hot line, IOD judges short message port, if client's hot line port is handed over client note a short message service Room; The short message service Room judges short message content, if stereotyped command, by the short message service Room, carry out normal process, if not stereotyped command, special interface command word will be used in the short message service Room, in interface protocol, regulation is 76003, by the Socket network of having set up, connect, this note is passed to the treatment system of the complicated note in the mobile client hot line short message service Room, the treatment system of the complicated note in the mobile client hot line short message service Room receives after this note, content is understood, then automatically replied or replied by operator;

Natural language processing module, for client's short message content is carried out to text preliminary treatment, identification and understanding, therefrom draw the structured message that consulting is relevant, described consulting relevant information comprises Business Name, theme, the summary that client seeks advice from, be called natural language processing result, i.e. NLP result; Simultaneously, according to system environments, information and natural language processing result that active user is relevant, determine whether to automatically reply, the NLP result automatically replying deposits in and automatically replies NLP outcome pool, otherwise deposits operator NLP outcome pool in;

Automatically reply module, realized automatically replying of short message content, do not need manual intervention; Automatically replying module comprises: obtain natural language processing object module, answer search module, answer template instances module and trigger action module, database access module;

First be to automatically reply module to obtain to meet and automatically reply tactful NLP result, here refer to that those are from the result of " automatically replying NLP outcome pool ", we are called the NLP result that can automatically reply, and then call answer template corresponding to answer search module automatic search NLP result; Then call answer template instances module and complete answer template instances; Finally, while comprising the business action of needs execution in answer, call the execution of trigger action module finishing service action, call answer simultaneously and reply module, concrete answer is replied to client; Like this, just automatically completed the processing of current consulting, in the process automatically replying, will produce client and seek advice from history, these consulting history will be saved in knowledge base by calling data storehouse access modules; Automatically reply the database access operation of the module in module, all to realize by calling data storehouse access modules, so also realized the multiplexing of module, automatically reply module and related to answer search module, answer template instances module and trigger action module, these modules are replied in module also called operator, so the public module of independent one-tenth, realizes the multiplexing of module;

Operator replys module, introduced this artificial role of operator the NLP result that can not automatically reply has been carried out to manual intervention processing, operator replys the modular structure of module, mainly comprise: natural language processing result distribution module, this natural language processing result refers to the NLP result that can not automatically reply; Intelligent answer module, a problem of this module input, directly calls NLP module and obtains summary corresponding to problem; Natural language processing result acquisition module; Obtain answer formwork module; Answer template instances module and trigger action module, trigger action module is called the trigger action module of the treatment system of the complicated note in the whole mobile client hot line short message service Room, completes the execution of concrete business action; Database access module;

First be the distribution of natural language processing result: the NLP result that can not automatically reply is assigned in corresponding consulting pond, operator can only be under with it gets consulting in those consulting ponds corresponding to technical ability group; Natural language result distribution module flow process is divided two steps, and the first step is " building consulting pond ": according to the configuring condition of operator's technical ability group in knowledge base, and initialization consulting pond; Second step will take out NLP result from " NLP outcome pool ", then according to the channel comprising in NLP result, districts and cities and brand message, obtain consulting pond corresponding to NLP result, finally this NLP result be put into consulting pond; This step can repeat always, constantly NLP result is put into consulting pond; Wherein, described NLP result distribution module is also distributed according to the service priority level of NLP result, and the service priority level of described consulting comprises from low to high successively: normal client rank, white list levels of clients, the rank that requests help, modification level; Wherein, in the same situation of the time of a plurality of consultings, the preferential higher consulting of distribution services priority level

Operator's natural language processing result acquisition module flow process: each operator belongs to one or more operator's technical ability groups, after operator logs in, just can determine consulting pond corresponding to these technical ability groups that he can be under him and get traffic; Like this, just determined that an operator is to the dictionary of seeking advice from pond array; First operator goes to obtain consulting from seek advice from that maximum consulting ponds, and the consulting is here a NLP result namely, and wherein consulting represents that at most this consulting pond is in the treatable consulting of this operator pond " the busiest "; Operator reaches consulting from its treatable that the busiest consulting pond, owing to there being a plurality of operators to obtain at the same time consulting, so when existing this operator to obtain less than consulting, obtain consulting from its treatable busy consulting pond, until it gets consulting; When operator gets consulting, operator replys the page will demonstrate the consulting that it gets;

Wherein, described operator replys module and also comprises consulting history buffer module and focus statistics module;

Described consulting history buffer module is for depositing the consulting historical data in setting-up time, described consulting historical data comprises consultation information and return information, described consultation information mainly comprises consultation service title, theme and summary, and described return information comprises replys Business Name, theme and summary and corresponding answer;

Described focus statistics module is for counting the consulting focus in certain hour, and described consulting focus replied to the hot zone that module is presented on display interface by operator and for operator, use;

Wherein, described operator replys module also for identification " reply type " automatically, replys type answertype and is defined as follows:

Answertype=0: represent that NLP has result, and operator confirms that NLP result is correct;

Answertype=1: represent that NLP has result, but operator confirms mistake, and process in the answer automatically being generated by system on basis and modify, send at NLP;

Answertype=2: represent that NLP has result, but operator confirms mistake, and pick up answer in knowledge base cache module, send;

Answertype=3: represent that NLP has result, but operator confirms mistake, and do not find answer in knowledge base cache module, from edlin answer, send;

Answertype=4: represent that NLP is without result, but operator finds answer to send in knowledge base cache module;

Answertype=5: represent that NLP is without result, but operator finds answer in knowledge base cache module, and revise in answer, send;

Answertype=6: represent that NLP is without result, and operator do not find answer in knowledge base cache module,, sent from edlin answer by operator;

Trigger action module is the triggering for finishing service action, the execution flow process of trigger action: the short message service Room, when not treatable note is given to the treatment system of the complicated note in the mobile client hot line short message service Room by it, also provides concrete action command data and the interface of finishing service action; First, trigger action module is set up Socket network by the interface providing with the short message service Room and is connected; Then trigger action module becomes the needed packet in the short message service Room by action command data encapsulation, uses the interface command word 70001 of agreement regulation, by the connection of setting up, the packet that comprises action command is pushed to the short message service Room; The short message service Room receives the packet that we send, and unpacks processing; Then to action command, whether be the judgement of stereotyped command, if so, according to normal flow, process the instruction that this client sends, and the note that the result of instruction is generated sends to client; Otherwise, send the SMS Tip of " instruction is wrong " to user;

Answer search module, for the search to mobile answer; Be mainly to provide NLP object information or " business+theme+summary " information, by the search of knowledge base and reasoning, obtain the answer that NLP result or " business+theme+summary " are corresponding;

Knowledge base searching module, for the search to mobile answer, but different with answer search module, the input of knowledge base searching module is the word of natural language, can think a complete problem content; Knowledge base searching module is by Automatically invoked natural language processing module, answer search module, and a plurality of optional answer to the correspondence that goes wrong according to the confidence level of understanding, and for operator, therefrom selects;

Answer template instances module, for the instantiation to knowledge templet, the answer of safeguarding in KBM module is one may comprise the stencil-like answer that business performs an action; This answer, need to, after instantiation, just can become a concrete answer; In answer template, the Template Information comprising comprises: Subscriber Number, brand, districts and cities, content, business, theme, summary, network address summary, trigger action summary, also comprise trigger action in answer template;

Module is replied in answer, for answer being replied to the client who sends consulting, according to the difference of channel, the reply mode of answer is different, the answer of note channel is replied by short message mode, the reply of BBS channel, the mode of leaving a message by BBS is replied, the reply of QQ channel, simulate QQ and send message, the to the effect that answer of correspondence of this message, the reply of MSN channel, simulates MSN and sends message;

Knowledge base maintenance module, for the maintenance management work of mobile service knowledge, comprises knowledge base user management module, region administration module, channel management module, customer type administration module, service management module, answer administration module;

Described knowledge base user management module comprises new user's registration, the statistics printing function of user's inquiry, modification, deletion and user profile form;

Described region administration module, for the effective management to the Regional Property of business;

Described channel management module, comprises the management to BBS, WAP, note, 10086 channels; Knowledge base maintenance engineer can pass through this module, safeguards " add to revise and delete " treatable electronic channel of system;

Described customer type administration module, comprises individual, family, the management of group customer type;

Described service management module, comprises the loading of business tree and the management of business-subject summary data; Between the business loading, there is hierarchical relationship, be convenient to browsing of contact staff, and knowledge base editorial staff's management; Knowledge base editorial staff selects any business, can load the data of choosing business; As a kind of embodiment, unreal-activity business and industry business under top layer, have successively been divided again; Professional knowledge administration module, manages with the data of traffic aided for knowledge base editorial staff couple;

Wherein, the data with traffic aided, comprising: " brand ", " business ", " theme ", " summary " and " answer " data.

Described service management module, also comprises:

Business Name editor module, for realizing the management of knowledge base editorial staff to Business Name, comprises increase, deletion, revises the structure of Business Name and modification business tree; Increase, delete and revise the full detail of mobile service;

Business Name checking module, for the input checking to Business Name and business structure is provided, makes Business Name meet the consistency constraint with customer type;

Theme summary editor module, for realizing the management of knowledge base editorial staff to business-subject and summary, comprises theme and the summary of increase, deletion, modification business;

Theme summary checking module, for the input checking to input subject name and summary title, comprises the consistency constraint of summary title and Business Name, the consistency constraint of summary, theme and brand;

Described consistency constraint, refers to the rule of the predefined interpolation data that meet system user requirement,

(1) if business ends up with " class " word, must add under " all clients " type;

(2) if business does not end up with " class " word, can not add under " all clients " type.

Because the hierarchical relationship of knowledge base itself is complicated, knowledge base is carried out to fine-grained cutting, adopt the representation of stratification, at the bottom of stratification, adopt the method for metadata automatic clustering, these metadata are carried out to Local Clustering, then the class of polymerization is carried out to new merging, reach according to this effect of Optimum Classification;

Answer administration module, carries out automatic clustering and management for the answer that knowledge base editorial staff is edited according to default level.

The answer that described answer administration module is edited knowledge base editorial staff is carried out automatic clustering and management according to the level of " term of validity=> number bag=> channel=> answer ", wherein:

The term of validity: when mobile subscriber or selected certain summary, what first show is the term of validity of all answers, and system provides the increase to the term of validity, and inquiry is revised and delete function.Choose wherein random time section, show the number bag under answer under this time period;

Number bag: system provides uploading of the bag of checking numbers equally, revises and delete function; Upper mark code is bundled into after merit, shows size and the routing information of respective number bag, and prompting user further checks;

Channel: comprise note, website, Fetion, BBS and WAP;

At present available is note channel, and note channel provides the short message service Room and two configurations of port; If have answer under note channel, show the short message service Room and port arrangement that answer is corresponding, and can re-start configuration;

Answer: KBS carries out automatic clustering according to the content of answer and city, place;

Knowledge base editorial staff can edit separately and delete the answer in each city, also can whole answer be modified and be deleted;

Described answer template knowledge management module, comprising: city rechecking module, and for retrieving the city that has answer.

A processing method for the complicated note in the mobile client hot line short message service Room, comprises the following steps:

The step 100. short message service Room is as client, and to the mobile client hot line short message service Room, complicated note system sends connection request, sets up Socket and connects;

Step 200. is accepted the complicated note that the short message service Room can not be understood it and process, and by current connection, sends to the complicated note system in the mobile client hot line short message service Room;

Step 300. natural language processing module will be responsible for the preliminary treatment of short message content, identification and understanding;

Step 310. automatically replies module by being responsible for very high to the accuracy of natural language processing result or meeting the NLP result that other automatically reply standard, automatically replies processing; Specifically: after the natural language processing result that automatic acquisition can automatically reply, call answer search module, search answer template corresponding to NLP result, then carry out the instantiation of answer template, may call trigger action module, help cellphone subscriber to obtain required answer or complete the business action that certain user requires simultaneously;

Step 320. operator signs in to traffic by login module and replys module, the NLP result not automatically replying is processed to reply, detailed process: operator judges natural language processing result, if NLP result is inaccurate, operator calls knowledge base searching or answer search module, search out the required correct option template of customer problem, then call answer template instances module or trigger action module, help cellphone subscriber to obtain required answer or complete concrete business action.Operator also can reply module by traffic treated consulting is reissued to processing;

Step 400. automatically replies module and operator replys module by calling answer search module, searches out the answer template of the selected business of NLP result or live traffice person, theme, summary correspondence in knowledge base;

The answer template that step 410. searches out, by answer template instances module, realizes the instantiation of answer template, generates concrete answer of replying, and then calls answer and replys the reply that module is carried out answer;

Step 420. in the answer template searching out, when comprising the performing an action of business, by trigger action module, completes the execution to business action;

The consulting of step 510. pair different channels comprises client's consulting of note channel, carries out the reply of corresponding answer.Beneficial effect:

1. the present invention can not understand the client note consulting feature (command type, knowledge type, data class etc.) of processing according to the short message service Room, sets up client's consulting model, can carry out better natural language processing.

2. the present invention forms client to customer knowledge demand classification summary and seeks advice from knowledge point, thereby customer service content is carried out to active planning, and fast the real information of client is passed to relevant departments of company, contribute to all departments of company to commence business in time and service optimizes, and guide the design and implementation of new product, business, really bring into play information center and the transducer effect of Customer Service Center;

3. the present invention has adopted existing extensive knowledge base modeling and architecture technology, can realize the centralized management of client's consulting and reply answer, voice and text have been realized in the fusion of bottom data stream, reached client's demand for counseling full dose, added up in real time, accurately and sort, realized the consulting of customer demand and the Intelligent Matching of information on services.

4. the consulting that the present invention " does not answer the short message service Room ", becomes the consulting of " can answer ".To most consultings, can directly by Intelligent Matching, realize and automatically replying.For the note that can not automatically reply processing, adopt the mode of " do and select True-False " to allow contact staff intervene processing, to process simple and conveniently, efficiency is high.

5. the present invention has set up contact between client and mobile service OSS (Business Operations Support System is called for short BOSS), can understand on the basis of customer demand, helps the execution of client's finishing service action.

accompanying drawing explanation will

Fig. 1 is the modular structure schematic diagram of system of the present invention;

Fig. 2 is that in the present invention, the flow chart of this treatment system is given to its not treatable note in the short message service Room;

Fig. 3 is the structure chart that automatically replies module in the present invention;

Fig. 4 is the structure chart that in the present invention, operator replys module;

Fig. 5 is the natural language result distribution module flow chart in the present invention;

Fig. 6 is the operator's natural language processing result acquisition module flow chart in the present invention;

Fig. 7 is the traffic reply page that in the present invention, operator replys module;

Fig. 8 is the short message service Room state configuration page in the present invention;

Fig. 9 is the flow chart of business trigger module finishing service trigger action implementation in the present invention;

Figure 10 is the stratification demonstration figure of knowledge base in the present invention;

Figure 11 is the demonstration figure of answer administration interface in knowledge base maintenance module in the present invention;

Figure 12 is the flow chart of the inventive method.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in more detail.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.

For can clearer explanation the present invention, below in conjunction with feature of the present invention and design, need the term being explained as follows:

1. note: Short Message Service, be called for short SMS, be that user passes through word or the digital information that mobile phone or other mobile terminals directly send or receive.

2. mobile client hot line: be to solve variety of issue and the difficulty that client runs in using mobile phone process, mobile operator has been set up customer service hotline throughout the country.Client can dial at any time local hot line number consulting, handle miscellaneous service, grasps the telephone expenses situation of oneself, understands the knowledge relevant with mobile communication, and mobile operator is presented one's view, advised and complains.

3. the short message service Room: be called for short the short Room, be one and offer the self-help serving system that client uses, the mobile phone client of mobile operator can by sending, the mode of note be carried out telephone expenses inquiry and business handling, and current preferential policy, the advertising campaign that also can provide company seeked advice from and handled etc.The short message service Room is the important component part of mobile operator electronic channel.

4. consulting: in Traffic handling system, the problem that certain client proposes, is called as a consulting.

5. traffic: various electronic channels are as the solution process of client's consulting in the sources such as note, wap, network, hotline.

6. operator: the staff that client's consulting of various channels is processed.As customer service hot line operator, be mainly to answer caller client, solve related service consulting, be engaged in all kinds of consoles such as toll traffic on duty, international traffic, directory enquiry, radio call, information service, subscriber exchange, and the personnel of service inquiry on processor.

7. operator's technical ability group: be called for short technical ability group.Operator's technical ability group is comprised of operator, and the technical ability that has represented operator is that these operators can process the traffic of which " channel districts and cities brand ".An operator can add a plurality of technical ability groups.

8. mobile knowledge: the knowledge that mobile service is relevant, mainly comprises the knowledge such as brand, business, theme, summary, answer.Introduce respectively below:

A. brand: brand herein specially refers to mobile difference of customer service being divided according to different types of service and so on.As China Mobile is divided into M-ZONE, walk in the Divine Land etc. by its main customer type.

B. business: business herein specially refers to the service name that the commmunication company of mobile field provides, and has hierarchical relationship between business.As the inquiry of M value, SIM under below M-ZONE serve etc.

C. theme: by the kind of PROBLEM DECOMPOSITION, as introduce class, use class etc.The operation that can carry out for specific business.As can be inquired about for M value inquiry business, canceling method etc.

D. make a summary: the abbreviation of answer summary knowledge,, be the concise and to the point narration of the standard advisory under specific theme.As for introducing class theme, the summary of its lower jurisdiction may comprise " introduction of M value business " etc.

E. answer: the answer that typical problem is corresponding, this answer is generally an answer template, that is to say, this answer is an answer that need to be specific instantiation.In answer, may also comprise performing an action of business.

For example: suppose " introduction of M value business " this summary under business " M value ", answer comprises: answer content is: " you are good for distinguished user; thank you to use <% Subscriber Number %> inquiry <% business %> business, China Mobile.", answer performs an action as " JSM ".

When user has sent note problem with phone number 13000000000: " what Dong Dong is M value business be? "

After natural language processing, the answer template obtaining: " you are good for distinguished user, thank you to use <% Subscriber Number %> inquiry <% business %>, China Mobile.”

Specialize and to obtain concrete answer after answer template: " you are good for distinguished user, thank you to use 13000000000 inquiry M value business, China Mobile.”

The action of carrying out: " JSM "

As shown in Figure 1, the treatment system of the complicated note in the mobile client hot line short message service Room comprises login module, note access module, natural language processing module, automatically replies module, operator replys module, trigger action module, answer search module, knowledge base searching module, answer template instances module, knowledge base maintenance module;

Introduce in detail the composition of modules below:

login module,be used for verifying log-on message.Operator by authentication after, enter operator and reply module.System user by authentication after, enter knowledge base maintenance module.According to the role's of system user difference, be divided into following several situation:

2. mobile contact staff logins by individual job number and password, by entering movable service knowledge base after authentication.

note access module, be the interface module of the short message service Room and system of the present invention.The flow process that Fig. 2 has described the not treatable note access native system of customer service hot line and processed, is first that client sends note to client's hot line, and IOD judges short message port, if client's hot line port is handed over client note a short message service Room.The short message service Room judges short message content, if stereotyped command, by the short message service Room, carry out normal process, if not stereotyped command, the short message service Room, by using special interface command word (in interface protocol, regulation is 76003), connects by the Socket network of having set up, this note is passed to our system, native system receives after this note, and content is understood, and then automatically replies or is replied by operator.

natural language processing module,for client's short message content is carried out to text preliminary treatment, identification and understanding, therefrom draw the structured message that consulting is relevant, described consulting relevant information comprises Business Name, theme, the summary that client seeks advice from, and is called natural language processing result (NLP result).Simultaneously according to system environments (native system can be directly and some third party's system communicate, complete some specific business function.As mobile business support system (this system can complete the operation of handling of all mobile services, is called for short BOSS system), by with the communicating by letter of this system, the execution that finishing service moves is indirectly as " opening 30 yuan of M-ZONE set meals ".The state set of the third party's system that here all native systems can be communicated by letter is called system environments.In the time of as the normal work of BOSS system, system environment variable BOSS system mode is set to " opening ", when BOSS system breaks down, system environment variable BOSS system mode is set to " pass "), information (user-dependent information that active user is relevant, mainly comprise phone number, the number bag (set of numbers) at the districts and cities' information under phone number and brand message, number place etc.) and natural language processing result, determine whether to automatically reply.The NLP result automatically replying deposits in and automatically replies NLP outcome pool, otherwise deposits operator NLP outcome pool in.The policy calculation that automatically replies wherein comprises: (equaling word mould is the special word mould of a class to equal word mould coupling, once matching, this class word mould just represents that this NLP result can automatically reply), generic word mould matching rate very high (higher than certain threshold value), class consulting at night (generally, within certain time period at night, do not have in operator's situation on duty, need to the traffic in this period be automatically replied, the content that needs that the time period of class at night is set and automatically reply user, all consultings that interior native system is received during this period of time are all called class consulting at night, they will be automatically replied), refuse messages class consulting (some short message content, may be insignificant mess code (short message content is all mess code character), or there is attack (in one section of official hour, the note quantity that certain number sends reaches the number of threshold values of regulation, we can assert it is to attack note) etc. note, for these notes, we think that they belong to refuse messages, system will automatically reply), the consulting of empty template class or other special defects consultings etc.

automatically reply module, realized automatically replying of short message content, do not need manual intervention (not needing operator's intervention to process).Fig. 3 has described modular structure and the invoked procedure that automatically replies module.Automatically replying module comprises: obtain natural language processing result (the NLP result that can automatically reply) module, answer search module, answer template instances module and trigger action module, database access module;

First be to automatically reply module to obtain to meet and automatically reply tactful NLP result and (refer to that those are from the result of " automatically replying NLP outcome pool " here, we are called the NLP result that can automatically reply), then call answer template corresponding to answer search module automatic search NLP result (answer of depositing in knowledge base is an answer template that need to be specific instantiation).Then call answer template instances module and complete answer template instances; Finally, while comprising the business action of needs execution in answer, call the execution of trigger action module finishing service action, call answer simultaneously and reply module, concrete answer is replied to client.Like this, just automatically completed the processing (we claim the processing of once seeking advice to be called traffic one time, and consulting can be automatically replied (automatically processing), also can manually be replied (operator replys processing)) of current consulting.In the process automatically replying, will produce client and seek advice from history, these consulting history will be saved in knowledge base by calling data storehouse access modules.Automatically reply the module that relates in module as answer search module etc., their database access operation, is all to realize by calling data storehouse access modules, has so also realized the multiplexing of module.Automatically reply module and related to answer search module, answer template instances module and trigger action module, these modules are replied in module also called operator, so the public module of independent one-tenth realizes the multiplexing of module.

operator replys module, introduced this artificial role of operator the NLP result that can not automatically reply carried out to manual intervention processing.Fig. 4 has described the modular structure that operator replys module, mainly comprise: natural language processing result (the NLP result that can not automatically reply) distribution module, an intelligent answer module (problem of this module input, directly call NLP module and obtain summary corresponding to problem), natural language processing result (the NLP result that can not automatically reply) acquisition module, obtain answer formwork module, (this module is called the trigger action module of whole system for answer template instances module and trigger action module, complete the execution of concrete business action), database access module.

First be the distribution of natural language processing result: the NLP result that can not automatically reply is assigned in corresponding consulting pond, operator can only be under with it gets consulting (from doing to the spare time) in those consulting ponds corresponding to technical ability group (one or more).Fig. 5 is the natural language result distribution module flow chart in the present invention, this flow process is divided two steps, the first step is " building consulting pond ": according to the configuring condition of operator's technical ability group in knowledge base, initialization consulting pond (each operator's technical ability group, the information that is comprising some channel, some districts and cities and some brand, corresponding represent that this technical ability group can process client's consulting of these channels, these districts and cities and these brands, according to technical ability group in these districts and cities, channel, brand message, carry out the set of initialization consulting pond.Wherein, channel+Yi Ge districts and cities+mono-brand, a corresponding unique consulting pond, consulting pond is just named with " channel+districts and cities+brand ".Such as channel=" note ", districts and cities=" Guangzhou ", brand=" M-ZONE ", a just corresponding unique consulting pond, the title in this consulting pond is exactly " note Guangzhou M-ZONE ".A technical ability group correspondence one or more consulting ponds).Second step will take out NLP result from " NLP outcome pool (the NLP result that can not automatically reply) ", then according to the channel comprising in NLP result, districts and cities and brand message, obtain consulting pond corresponding to NLP result, finally this NLP result be put into consulting pond.This step can repeat always, constantly NLP result is put into consulting pond.Wherein, described NLP result distribution module is also distributed according to the service priority level of NLP result, and the service priority level of described consulting comprises from low to high successively: normal client rank, white list levels of clients, the rank that requests help, modification level.Wherein, when a plurality of consultings arrive in the same situation of time of system of the present invention, the preferential higher consulting of distribution services priority level.In Fig. 5, suppose in knowledge base: the technical ability group that has three brands of " Guangzhou " districts and cities (M-ZONE, Global Link, walk in the Divine Land) that can process " note " channel, the result that builds technical ability group is: by these technical ability groups, built three consulting ponds " note Guangzhou M-ZONE ", " note Guangzhou Global Link ", " walk in the Divine Land of note Guangzhou " three consulting ponds.

Fig. 6 is the operator's natural language processing result acquisition module flow chart in the present invention.Each operator belongs to one or more operator's technical ability groups, after operator logs in, just can determine consulting pond (may be a plurality of) corresponding to these technical ability groups that he can be under him and get traffic.Like this, just determined that an operator is to the dictionary of seeking advice from pond array.For example, " operator's technical ability group A " can process the traffic of three brands of " Guangzhou " districts and cities (M-ZONE, Global Link, walk in the Divine Land) of " note " channel, and " operator's technical ability group B " can process the traffic of three brands of " Dongguan " districts and cities (M-ZONE, Global Link, walk in the Divine Land) of " note " channel.Operator C belongs to operator's technical ability group A and operator's technical ability group B, at " operator-consulting pond relation " dictionary, added keyword for " operator C ", corresponding consulting pond array is { consulting pond " note Guangzhou M-ZONE ", consulting pond " note Guangzhou Global Link ", consulting pond " walk in the Divine Land of note Guangzhou ", consulting pond " note Dongguan M-ZONE ", consulting pond " note Dongguan Global Link ", consulting pond " walk in the Divine Land of note Dongguan " }.First operator goes to obtain consulting from seek advice from that maximum consulting ponds, and (the consulting is here a NLP result namely, and wherein consulting represents that at most this consulting pond is in the treatable consulting of this operator pond " the busiest ".Operator reaches consulting from its treatable that the busiest consulting pond, owing to there being a plurality of operators to obtain at the same time consulting, so when existing this operator to obtain less than consulting, obtain consulting from its treatable busy consulting pond, until it gets consulting).When operator gets consulting, operator replys the page will demonstrate the consulting that it gets.

Fig. 7 is exactly that the traffic that in the present invention, operator replys in module is replied the page, and the user profile wherein getting comprises: districts and cities=" Shaoguan ", brand=" M-ZONE ", channel=" website ", phone number=" 15907513731 ".The problem of consulting=" my telephone expenses remaining sum also has how many? " NLP result comprises: business=" inquiry telephone expenses ", theme=" use ", summary=" telephone expenses inquiry into balance method ", the concrete answer after instantiation=" <p> inquiry telephone expenses remaining sum.Open in the short Room.</p> ", the port of answer short message=" 10086 ".

Fig. 8 is the short message service Room state configuration page in the present invention.(short message service Room state belongs to a part for system environments)

Wherein, described operator replys module and also comprises consulting history buffer module and focus statistics module.

Described focus statistics module is for counting the consulting focus in certain hour, and described consulting focus replied to the hot zone that module is presented on display interface by operator and for operator, use.

Wherein, described operator replys module also (because NLP result is not necessarily correct, so need operator's artificial judgment, at this moment just needs to use " reply type " this concept for identification " reply type " automatically.Reply type and can help knowledge base maintenance personnel to improve the quality of knowledge base, thereby improve the precision of natural language processing), reply type answertype and be defined as follows:

Answertype=6: represent that NLP is without result, and operator do not find answer in knowledge base cache module,, sent from edlin answer by operator.

trigger action moduleit is the triggering for finishing service action.Fig. 9 has described the execution flow process of trigger action, and the short message service Room, when not treatable note is given to the treatment system of the complicated note in the mobile client hot line short message service Room by it, also provides concrete action command data and the interface of finishing service action.First, trigger action module is set up Socket network by the interface providing with the short message service Room and is connected; Then trigger action module becomes the needed packet in the short message service Room by action command data encapsulation, uses the interface command word 70001 of agreement regulation, by the connection of setting up, the packet that comprises action command is pushed to the short message service Room.The short message service Room receives the packet that we send, and unpacks processing.Then to action command, whether be the judgement of stereotyped command, if so, according to normal flow, process the instruction that this client sends, and the note that the result of instruction is generated sends to client.Otherwise, send the SMS Tip of " instruction is wrong " to user.

Like this, for user, when he is when send carrying out note that certain business action request is content to the short message service Room, if the short message service Room can not be processed, it also can be given to note system of the present invention, is processed, then be given to the triggering of short message service Room execution by system of the present invention, expand the function in the short message service Room, strengthened the ability of its natural language processing.Simultaneously, when client sends the note of carrying out certain business action request to time our system, we also can help the execution of user's execution automatically at system, like this, also expand the function of our systems, strengthened the executive capability of the business action of our systems.

answer search module, for the search to mobile answer; Be mainly to provide NLP result (or business+theme+summary) information, by the search of knowledge base and reasoning, obtain answer corresponding to NLP result (or business+theme+summary).

knowledge base searching module, for the search to mobile answer, but different with answer search module, the input of knowledge base searching module is the word of natural language, can think a complete problem content.Knowledge base searching module is by Automatically invoked natural language processing module, answer search module, and a plurality of optional answer to the correspondence that goes wrong according to the confidence level of understanding, and for operator, therefrom selects.

answer template instances module, for the instantiation to knowledge templet.The answer of safeguarding in KBM module is one may comprise the stencil-like answer that business performs an action.This answer, need to, after instantiation, just can become a concrete answer.(in answer template, the Template Information that can comprise comprises: Subscriber Number, brand, districts and cities, content, business, theme, summary, network address summary, trigger action summary etc., also may comprise trigger action in answer template.）

module is replied in answer,for answer being replied to the client who sends consulting.According to the difference of channel, the reply mode of answer is different.The answer of note channel is replied by short message mode, the reply of BBS channel, and the mode of leaving a message by BBS is replied.The reply of QQ channel, simulates QQ and sends message (the to the effect that answer of correspondence of this message), and MSN similarly.

knowledge base maintenance module, for the maintenance management work of mobile service knowledge, comprise knowledge base user management module, region administration module, channel management module, customer type administration module, service management module, answer administration module;

Described knowledge base user management module comprises new user's registration, the statistics printing function of user's inquiry, modification, deletion and user profile form.

Described region administration module, for the effective management to the Regional Property of business.All business regional informations form one tree, and we are called region tree, if Figure 10 is the stratification demonstration figure of knowledge base in the present invention.

Described channel management module, comprises the management to BBS, WAP, note, the channels such as 10086.Knowledge base maintenance engineer can pass through this module, safeguards (add to revise and delete) treatable electronic channel of system.

Described customer type administration module, comprises individual, family, the management of the customer types such as group.

Described service management module, comprises the loading of business tree and the management of business-subject summary data.Between the business loading, there is hierarchical relationship, be convenient to browsing of contact staff, and knowledge base editorial staff's management.Knowledge base editorial staff selects any business, can load the data of choosing business.As a kind of embodiment, in the present invention, the available service of top layer is " brand ", has successively divided again unreal-activity business (not having this business, only for dividing level) and industry business under " brand ".Professional knowledge administration module, manages with the data of traffic aided for knowledge base editorial staff couple;

Wherein, the data with traffic aided, comprising: the data such as " brand ", " business ", " theme ", " summary " and " answer ".

Described service management module, also comprises:

Theme summary checking module, for the input checking to input subject name and summary title, comprises the consistency constraint of summary title and Business Name, the consistency constraint of summary, theme and brand.

Described consistency constraint, refers to the rule of the predefined interpolation data that meet system user requirement, for example:

Due to the features such as hierarchical relationship complexity of knowledge base itself, so we have adopted fine granularity, multi-level knowledge base modeling and administrative skill.We are in the building process of this knowledge base, adopt various dimensions, omnibearing structure pattern, in conjunction with of mobile project itself, from " the city title " of higher discrimination, to " customer type ", be then " type of service ", knowledge base is carried out to fine-grained cutting, adopt the representation of stratification, at the bottom of stratification, adopt the method for metadata automatic clustering, these metadata are carried out to Local Clustering, then the class of polymerization is carried out to new merging, reach according to this effect of Optimum Classification.Such as three business " inquiry telephone expenses " for mobile, " inquiry ownership place " and " set meal has been opened in inquiry ", can be divided into " inquiry class " together by these three business.Adopting choice box and tree figure is a kind of good data structure that realizes multilayered structure.

Answer administration module, as shown in answer administrative template interface display figure in knowledge base maintenance module in Figure 11 the present invention, carries out automatic clustering and management for the answer that knowledge base editorial staff is edited according to default level.

As a kind of embodiment, the answer that in the present invention, answer administration module is edited knowledge base editorial staff is carried out automatic clustering and management according to the level of " term of validity=> number bag=> channel=> answer ", wherein:

The term of validity: when mobile subscriber or selected certain summary, what first show is the term of validity of all answers, and system provides the increase to the term of validity, and inquiry is revised and delete function.Choose wherein random time section, show the number bag under answer under this time period.

Number bag: system provides uploading of the bag of checking numbers equally, revises and delete function.Upper mark code is bundled into after merit, shows size and the routing information of respective number bag, and prompting user further checks.

Channel: comprise note, website, Fetion, BBS and WAP.

At present available is note channel, and note channel provides the short message service Room and two configurations of port.If have answer under note channel, show the short message service Room and port arrangement that answer is corresponding, and can re-start configuration.

Answer: KBS carries out automatic clustering according to the content of answer and city, place.

Knowledge base editorial staff can edit separately and delete the answer in each city, also can whole answer be modified and be deleted.

Described answer template knowledge management module, comprising:

City rechecking module for retrieving the city that has answer, only shows the city that does not configure answer on interface, prevents the interpolation that repeats of answer, guarantees data consistency.

Beneficial effect of the present invention is:

First, by building native system, be connected with mobile customer service hot line short message service Room system applies layer, not only can process its not treatable complicated client note and knowledge requirement class note, expand the existing function in the short message service Room, strengthened the natural language processing ability of short message service Room system and the ability of processing knowledge requirement class; And connect by this, short message service Room system can help native system to complete the desired business action of client.When the client of our system (directly sends problems to the client of our system by various electronic channels, as the client of direct transmission note to our system) when having the requirement of action executing, our system can help it to complete, this function has also been expanded the existing function of Intelligence repository, allows Intelligence repository indirectly have the ability of execution business action;

Second, the IKBS combining as construction of knowledge base, search and natural language processing, also by the present invention, realized and provide knowledge services for other system, for different channels, as unified reply of the consulting of the access waies such as phone, note, Fetion, WAP, QQ, WEB, the short Room, process, provide the foundation.Greatly improve the speed that employee searches business, processes traffic, improve staffing effectiveness, contribute to reduce employee's pressure simultaneously, reach the doulbe-sides' victory of inside and outside portion customer service satisfaction.

As shown in figure 12, for realizing object of the present invention, also provide the processing method of the complicated note in a kind of mobile client hot line short message service Room, comprise the following steps:

Connection request, as client, is sent to native system (the note access module in native system) in the step 100. short message service Room, sets up Socket and connects.

Step 200. is accepted the complicated note that the short message service Room can not be understood it and process, and by current connection, sends to native system;

Step 310. automatically replies module by being responsible for very high to the accuracy of natural language processing result or meeting the NLP result that other automatically reply standard, automatically replies processing; Specifically: after the natural language processing result that automatic acquisition can automatically reply, call answer search module, search answer template corresponding to NLP result, then carry out the instantiation of answer template, may call trigger action module, help cellphone subscriber to obtain required answer or complete the business action that certain user requires simultaneously.

The consulting of step 510. pair different channels comprises client's consulting of note channel, carries out the reply of corresponding answer.

Claims

1. a treatment system for the complicated note in the mobile client hot line short message service Room, is characterized in that: comprise login module, note access module, natural language processing module, automatically reply module, operator replys module, trigger action module, answer search module, knowledge base searching module, answer template instances module, knowledge base maintenance module;

Wherein, the data with traffic aided, comprising: " brand ", " business ", " theme ", " summary " and " answer " data;

Described service management module, also comprises:

(2) if business does not end up with " class " word, can not add under " all clients " type;

2. the treatment system of the complicated note in the mobile client hot line short message service according to claim 1 Room, is characterized in that:

The term of validity: when mobile subscriber or selected certain summary, what first show is the term of validity of all answers, and system provides the increase to the term of validity, and inquiry is revised and delete function;

Choose wherein random time section, show the number bag under answer under this time period;

Channel: comprise note, website, Fetion, BBS and WAP;

3. the processing method of the complicated note in the mobile client hot line short message service according to claim 1 Room, is characterized in that:

Comprise the following steps:

Step 320. operator signs in to traffic by login module and replys module, the NLP result not automatically replying is processed to reply, detailed process: operator judges natural language processing result, if NLP result is inaccurate, operator calls knowledge base searching or answer search module, search out the required correct option template of customer problem, then call answer template instances module or trigger action module, help cellphone subscriber to obtain required answer or complete concrete business action;

Operator also can reply module by traffic treated consulting is reissued to processing;