Summary of the invention
The application provides a kind of intelligent response method and apparatus, exists and passes through primary retrieval solving prior art
Method from storehouse, knowledge point, obtain the knowledge point mated with problem be difficult to improve and answer the asking of accuracy
Topic.The application additionally provides a kind of intelligent response system, and a kind of electronic equipment.
The application provides a kind of intelligent response method, it is characterised in that including:
Obtain corresponding specific the to be answered a question person of being asked and be evaluated as undesirable original answer;
With the information in described undesirable original answer for retrieval foundation, prestoring with predetermined standard
Knowledge point set in retrieve, with improvement answer to be answered a question described in obtaining;Described knowledge point includes mark
Topic and content.
Optionally, described undesirable original answer includes title;
Described predetermined standard includes: the highest with the title similarity of described undesirable original answer.
Optionally, if the knowledge point the highest with the title similarity of described undesirable original answer has many
Individual, then select to update the nearest knowledge point of time gap current time as described improvement answer.
Optionally, described undesirable original answer includes title;
Described predetermined standard includes:
Find out the knowledge point the highest with the title similarity of described undesirable original answer to know as candidate
Know point;
Judge that the URL comprised in described candidate knowledge point is undesirable the most former with described
Beginning answer is identical;
The most then get rid of this candidate knowledge point, find out, in residue knowledge point, the knowledge that title similarity is the highest
Point is as described candidate knowledge point, and returns previous judgement step;
If it is not, then using described candidate knowledge point as described improvement answer.
Optionally, described predetermined standard includes: with in the content of described undesirable original answer
Paragraph heading similarity is the highest.
Optionally, if the highest with the paragraph heading similarity in the content of described undesirable original answer
Knowledge point have multiple, then select to update the nearest knowledge point of time gap current time and answer as described improvement
Case.
Optionally, described predetermined standard includes:
Find out the knowledge the highest with the paragraph heading similarity in the content of described undesirable original answer
Point is as candidate knowledge point;
Judge that the URL comprised in described candidate knowledge point is undesirable the most former with described
Beginning answer is identical;
The most then get rid of this candidate knowledge point, in residue knowledge point, find out the paragraph heading in content similar
Spend the highest knowledge point as described candidate knowledge point, and return previous judgement step;
If it is not, then using described candidate knowledge point as described improvement answer.
Optionally, described predetermined standard includes: according to the title with described undesirable original answer
Paragraph heading similarity in similarity, content and uniform resource locator similarity, calculate overall similarity,
Using described overall similarity with the described undesirable immediate knowledge point of original answer as improving knowledge
Point;Described overall similarity employing formula calculated as below:
W=X*N1+Y*N2 Z*N3
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator.
Optionally, described predetermined standard includes:
Judge that current time is the most default less than or equal to first with the time interval of the recent renewal time of knowledge point
Time interval;
If above-mentioned judged result is yes, then according to the title similarity of described undesirable original answer,
Paragraph heading similarity in content and uniform resource locator similarity, and described current time is with described
The time interval of the recent renewal time of knowledge point, calculates overall similarity, with described overall similarity and institute
State the undesirable immediate knowledge point of original answer as improving knowledge point;Described overall similarity is adopted
Use formula calculated as below:
W=X*N1+Y*N2 Z*N 3+N4* (Ts-t)
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator, when N4 is default
Between weight, Ts is described first prefixed time interval, t be described current time with described knowledge point the most more
The time interval of new time.
Optionally, described predetermined standard includes: only made the knowledge point updated in stipulated time threshold range
The scope of described improvement answer is obtained for retrieval;Described time threshold is current time with described knowledge point
The preset value of the nearly time interval length updating the time.
Optionally, it is characterised in that retrieve and obtain with predetermined standard in the knowledge point set prestored described
Before improvement answer to be answered a question described in taking, also include:
Calculate and store the described title similarity of any two knowledge point.
Optionally, according to the second prefixed time interval or when updating described knowledge point, calculate and store and appoint
The described title similarity of two knowledge points of meaning.
Optionally, the described title similarity of described calculating any two knowledge point includes:
Resolve the first knowledge point and the title of the second knowledge point, obtain described first knowledge point and the second knowledge point
The predefined key word that each includes of title, respectively as the first title keyword set and the second mark
Topic keyword set;
According to described first title keyword set and the second title keyword set, calculate described first knowledge
Point and the title similarity of the second knowledge point.
Optionally, described title similarity uses the calculating of following formula to obtain:
X (A, B)=| A ∩ B |/| A U B |
Wherein, A is described first title keyword set, and B is described second title keyword set, X (A,
B) it is described title similarity.
Optionally, wait to answer described in retrieval acquisition with predetermined standard in the knowledge point set prestored described
Before the improvement answer of problem, also include:
Calculate and store the paragraph heading similarity in the described content of any two knowledge point.
Optionally, according to the second prefixed time interval or when updating described knowledge point, calculate and store and appoint
Paragraph heading similarity in the described content of two knowledge points of meaning.
Optionally, the paragraph heading similarity in the described content of described calculating any two knowledge point includes:
Resolve the described paragraph heading of the first knowledge point and the second knowledge point, obtain described first knowledge point and the
The predefined key word that the described paragraph heading of two knowledge points each includes, by described first knowledge point
The key word that described paragraph heading includes is as the keyword set of the first paragraph heading, by described second knowledge
The key word that the described paragraph heading of point includes is as the keyword set of the second paragraph heading;
Keyword set according to described first paragraph heading and the keyword set of the second paragraph heading, calculate
Described first knowledge point and the paragraph heading similarity in the content of the second knowledge point.
Optionally, the paragraph heading similarity in described content uses the calculating of following formula to obtain:
Y (C, D)=| C ∩ D |/| C U D |
Wherein, C is the keyword set of described first paragraph heading, and D is the key of described second paragraph heading
Set of words, Y (C, D) is the paragraph heading similarity in described content.
Optionally, wait to answer described in retrieval acquisition with predetermined standard in the knowledge point set prestored described
Before the improvement answer of problem, also include:
Calculate and store the uniform resource locator similarity in the described content of any two knowledge point.
Optionally, according to the second prefixed time interval or when updating described knowledge point, calculate and store and appoint
Uniform resource locator similarity in the described content of two knowledge points of meaning.
Optionally, the uniform resource locator similarity in the described content of described calculating any two knowledge point
Including:
Resolve each content of the first knowledge point and the second knowledge point, obtain described first knowledge point and know with second
Know the uniform resource locator each included of point, respectively as the first uniform resource locator set and second
Uniform resource locator set;
According to described first uniform resource locator set and the second uniform resource locator set, calculate described
First knowledge point and the described uniform resource locator similarity of the second knowledge point.
Optionally, described uniform resource locator similarity uses the calculating of following formula to obtain:
Z (E, F)=| E ∩ F |/| E U F |
Wherein, E is described first uniform resource locator set, and F is described second uniform resource locator collection
Closing, Z (E, F) is described uniform resource locator similarity.
Optionally, wait to answer described in retrieval acquisition with predetermined standard in the knowledge point set prestored described
Before the improvement answer of problem, also include:
Calculate and store the described overall similarity of any two knowledge point.
Optionally, according to the second prefixed time interval or when updating described knowledge point, calculate and store and appoint
The described overall similarity of two knowledge points of meaning.
Optionally, described undesirable original answer uses following steps to generate:
Receive inquiry request to be answered a question described in the correspondence that client sends;
According to the described question and answer relation waited and answer a question and prestore, to be answered a question described in acquisition one or many
Individual original answer;
The original answer that quizmaster is chosen from the one or more original answer, as described not
Satisfactory original answer.
Optionally, described in described basis, wait the question and answer relation answered a question and prestore, described in acquisition, treat that answer is asked
The one or more original answer of topic, including:
Wait described in parsing to answer a question, the predefined key word that waiting described in acquisition answers a question includes;
According to described key word and described question and answer relation, obtain the answer with described Keywords matching, as institute
State one or more original answer.
Optionally, described knowledge point is structured document.
Accordingly, the application also provides for a kind of intelligent response device, it is characterised in that including:
Acquiring unit, is evaluated as undesirable for obtaining corresponding specific the to be answered a question person of being asked
Original answer;
Retrieval unit, for the information in described undesirable original answer for retrieval foundation, with in advance
Fixed standard is retrieved in the knowledge point set prestored, with improvement answer to be answered a question described in acquisition;Institute
State knowledge point and include title and content.
Optionally, described retrieval unit includes:
First retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
It is up to search criteria with the title similarity of described improvement answer Yu described undesirable original answer,
Retrieve in the described knowledge point prestored is gathered, with described improvement answer to be answered a question described in acquisition.
Optionally, described retrieval unit includes:
Second retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
Retrieve in the described knowledge point prestored is gathered, with described improvement answer to be answered a question described in acquisition;
Described second retrieval subelement includes:
First searches subelement, for finding out title similarity with described undesirable original answer
High knowledge point is as candidate knowledge point;
Judgment sub-unit, for judge the URL that comprises in described candidate knowledge point whether with institute
State undesirable original answer identical;
Second searches subelement, if the URL comprised in described candidate knowledge point is with described
Undesirable original answer is identical, then get rid of this candidate knowledge point, finds out title in residue knowledge point
The highest knowledge point of similarity is as described candidate knowledge point, and returns previous judgement step;
Judge subelement, if the URL comprised in described candidate knowledge point is not inconsistent with described
Close the original answer required to differ, then using described candidate knowledge point as described improvement answer.
Optionally, described retrieval unit includes:
3rd retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
The highest with the paragraph heading similarity in the content of described undesirable original answer with described improvement answer
For search criteria, retrieve in the described knowledge point prestored is gathered, described in be answered a question described in acquisition
Improve answer.
Optionally, described retrieval unit includes:
4th retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
Retrieve in the described knowledge point prestored is gathered, with described improvement answer to be answered a question described in acquisition;
Described 4th retrieval subelement includes:
First searches subelement, for finding out and the paragraph in the content of described undesirable original answer
The highest knowledge point of title similarity is as candidate knowledge point;
Judgment sub-unit, for judge the URL that comprises in described candidate knowledge point whether with institute
State undesirable original answer identical;
Second searches subelement, if the URL comprised in described candidate knowledge point is with described
Undesirable original answer is identical, then get rid of this candidate knowledge point, finds out content in residue knowledge point
In the highest knowledge point of paragraph heading similarity as described candidate knowledge point, and return previous judgement step;
Judge subelement, if the URL comprised in described candidate knowledge point is not inconsistent with described
Close the original answer required to differ, then using described candidate knowledge point as described improvement answer.
Optionally, described retrieval unit includes:
5th retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
It is up to search criteria with the overall similarity of original improvement answer Yu described undesirable original answer,
Retrieve in the described knowledge point prestored is gathered, with described improvement answer to be answered a question described in acquisition;
Described overall similarity be according to the title similarity of described undesirable original answer, content
In paragraph heading similarity and uniform resource locator Similarity Measure obtain;Described overall similarity is adopted
Use formula calculated as below:
W=X*N1+Y*N2 Z*N3
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator.
Optionally, described retrieval unit includes:
6th retrieval subelement, is used for the information in described undesirable original answer for retrieval foundation,
It is up to search criteria with the overall similarity of described improvement answer Yu described undesirable original answer,
Retrieve in the described knowledge point prestored is gathered, with described improvement answer to be answered a question described in acquisition;
Described overall similarity be according to the title similarity of described undesirable original answer, content
In paragraph heading similarity and uniform resource locator similarity, and described current time and knowledge point
The time interval of recent renewal time calculates acquisition;Described overall similarity employing formula calculated as below:
W=X*N1+Y*N2 Z*N3+N4* (Ts-t)
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator, when N4 is default
Between weight, Ts is the first prefixed time interval, when t is described current time with the recent renewal of described knowledge point
Between time interval.
Optionally, described device also includes:
First computation subunit, for calculating the described title similarity of any two knowledge point;
Described first computation subunit includes:
Resolve subelement, for resolving the first knowledge point and the title of the second knowledge point, obtain described first and know
Know the predefined key word that the title of point and the second knowledge point each includes, close respectively as the first title
Keyword set and the second title keyword set;
Computation subunit, is used for according to described first title keyword set and the second title keyword set,
Calculate the title similarity of described first knowledge point and the second knowledge point.
Optionally, described device also includes:
Second computation subunit, the paragraph heading in the described content calculating any two knowledge point is similar
Degree;
Described second computation subunit includes:
Resolve subelement, for resolving the first knowledge point and the described paragraph heading of the second knowledge point, obtain institute
State the predefined key word that the described paragraph heading of the first knowledge point and the second knowledge point each includes, will
The key word that the described paragraph heading of described first knowledge point includes is as the keyword set of the first paragraph heading
Closing, the key word included by the described paragraph heading of described second knowledge point is as the key of the second paragraph heading
Set of words;
Computation subunit, for according to the keyword set of described first paragraph heading and the second paragraph heading
Keyword set, calculates the paragraph heading similarity in the content of described first knowledge point and the second knowledge point.
Optionally, described device also includes:
3rd computation subunit, the unified resource location in the described content calculating any two knowledge point
Device similarity;
Described 3rd computation subunit includes:
Resolve subelement, for resolving the first knowledge point and each content of the second knowledge point, obtain described the
One knowledge point and the uniform resource locator each included of the second knowledge point, respectively as the first unified resource
Localizer set and the second uniform resource locator set;
Computation subunit, for positioning according to described first uniform resource locator set and the second unified resource
Device set, calculates the described uniform resource locator similarity of described first knowledge point and the second knowledge point.
Optionally, described device also includes:
4th computation subunit, for calculating the described overall similarity of any two knowledge point.
Optionally, also include:
Signal generating unit, is used for generating described undesirable original answer;
Described signal generating unit includes:
Receive subelement, for receiving inquiry request to be answered a question described in the correspondence that client sends;
Obtain subelement, be used for the question and answer relation waiting to answer a question and prestore described in basis, treat back described in acquisition
The one or more original answer of question and answer topic;
Choose subelement, original for that quizmaster is chosen from the one or more original answer
Answer, as described undesirable original answer.
Optionally, described acquisition subelement includes:
Resolve subelement, be used for resolving described in wait to answer a question, waiting described in acquisition to answer a question includes in advance
The key word of definition;
Inquiry subelement, for according to described key word and described question and answer relation, obtains and described key word
The answer joined, as the one or more original answer.
The application also provides for a kind of intelligent response system, including: according to the intelligent response described in any of the above-described item
Device.
Additionally, the application also provides for a kind of electronic equipment, described electronic equipment includes:
Display;
Processor;And
Memorizer, described memorizer is configured to store to be waited to answer a question, described in wait to answer a question by described place
When reason device performs, show at described display described in improvement answer to be answered a question, described improvement answer is
The information being evaluated as in undesirable original answer with the person of being asked is retrieval foundation, with predetermined standard
In the knowledge point set prestored, retrieval obtains, described original answer be according to described in wait to answer a question and in advance
The question and answer relation deposited generates.
Optionally, described knowledge point and described question and answer relation are stored in server end.
Optionally, described electronic equipment is wearable intelligent equipment;Described improvement answer is stored in this Wearable
In the memorizer of smart machine.
Compared with prior art, the application has the advantage that
Intelligent response method, device, system and the electronic equipment that the application provides, corresponding specific by obtaining
The to be answered a question person of being asked is evaluated as undesirable original answer, and with described undesirable
Information in original answer is retrieval foundation, retrieves in the knowledge point set prestored with predetermined standard, with
Improvement answer to be answered a question described in acquisition so that query and search to wait to answer a question the most corresponding
After answer, additionally it is possible to carry out quadratic search according to the undesirable original answer that quizmaster selects, obtain
Improvement answer to be answered a question.Owing to quizmaster reflects quizmaster to answer to the selection of original answer
Demand, can be as new retrieval foundation.Therefore, this method can be difficult to lead in prior art
In the case of crossing the information raising answer coverage rate and accuracy rate that waiting answers a question itself comprises, new by introducing
Retrieval according to improving the coverage rate of answer and accuracy rate.
Detailed description of the invention
Elaborate a lot of detail in the following description so that fully understanding the application.But the application
Can implement to be much different from alternate manner described here, those skilled in the art can without prejudice to
Doing similar popularization in the case of the application intension, therefore the application is not limited by following public being embodied as.
In this application, it is provided that a kind of intelligent response method and apparatus, and a kind of intelligent response system and
Electronic equipment.It is described in detail the most one by one.
The intelligent response method that the application provides, is based on such a design concept, it may be assumed that use question and answer knot
The mode that really secondary is recommended, after getting " once result " according to " problem " retrieval, it is possible to automatically lead to
Cross " once result " retrieval " second fruiting ".The intelligent response method provided due to the application is with " result
Result " relevant dimension supplement " problem result " single dimension, it is thus possible to promote solution problem
Coverage rate.
Refer to Fig. 1, it is the flow chart of intelligent response embodiment of the method for the application.Described method include as
Lower step:
Step S101: obtain corresponding specific the to be answered a question person of being asked and be evaluated as undesirable original
Answer.
Original answer described in the embodiment of the present application and improvement answer are described in the embodiment of the present application of knowledge point
Knowledge point have recorded solution user some or the flow process of a certain class problem, operating procedure, related system link
Etc. information.Knowledge point includes title and content.In actual applications, knowledge point can be a formatting literary composition
Shelves, such as: web document, word document etc.;Knowledge point can also is that unformatted document, such as: literary composition
Presents.Above-mentioned different knowledge point storage format, the most simply change of detailed description of the invention, all without departing from
The core of the application, the most all within the protection domain of the application.
In the present embodiment, each knowledge point is a web document.Refer to Fig. 2, it is the application
Intelligent response embodiment of the method in the schematic diagram of knowledge point.It can be seen in fig. 2 that a web document represents
Knowledge point include three parts: each paragraph heading in the title of knowledge, content, the unification in content
Resource localizer.Wherein, the title of knowledge represents the headline that this knowledge point is corresponding, and the title of knowledge also may be used
To include subtitle;Paragraph heading is the title of each ingredient of a knowledge point, it may be assumed that subhead;
The particular content of each ingredient of knowledge point is the concrete solution party being linked to by uniform resource locator
The information such as case.
The method that the application to be implemented provides, first has to, in corresponding specific answer to be answered a question, obtain
The person of being asked is evaluated as undesirable original answer.Generally, an intelligent response system is according to quizmaster
The problem proposed, in existing question and answer relation, retrieval obtains " the once result " of correspondence problem, it may be assumed that original
Answer.Owing to question and answer relation is the relation of one-to-many, the problem that therefore corresponding quizmaster proposes, can pass through
Intelligent response system obtains multiple original answers." once result " potentially includes multiple original answer, works as enquirement
After person has browsed these original answers, find, time answer is unsatisfied, to select to be thought general orientation
Correct (imagining closest with quizmaster) original answer notice intelligent response system, intelligent response system is again
The original answer that the solution problem general orientation specified according to this quizmaster is correct, obtains and improves answer.This Shen
The person of being asked described in embodiment please be evaluated as undesirable original answer, be that above-mentioned " quizmaster refers to
The original answer that fixed solution problem general orientation is correct ", but the particular content of this original answer can not be real
Solution problem.
In the present embodiment, the generation process of described original answer is divided into two steps: 1) receives client and sends out
Inquiry request to be answered a question described in the correspondence sent;2) wait described in basis that the question and answer answered a question and prestore are closed
It is, one or more original answer to be answered a question described in acquisition.
1) inquiry request to be answered a question described in the correspondence that client sends is received.
Client described in the embodiment of the present application includes the terminal unit such as PC, PAD, iPad, Yi Jiyi
Dynamic communication apparatus, it may be assumed that usually said mobile phone or smart mobile phone.The intelligent response method that the application provides,
The question and answer result secondary that its application scenarios is also not limited in online customer service system is recommended, and needs at other
Carry out according to once result to use method provided herein, such as under the scene of secondary recommendation:
If certain application is in order to improve the accuracy of solution, does not require nothing more than and retrieve once result according to problem,
Also need to the second fruiting according to the recommendation of once result with more high accuracy, in the field having similar Search Requirement
The method that the application can be applied under scape to provide, carries out question and answer result secondary and recommends.
2) according to described in wait the question and answer relation answering a question and prestore, to be answered a question described in acquisition one or
Multiple original answers.
Receiving correspondence that client sends after the inquiry request answered a question, answer a question according to waiting and
The question and answer relation prestored, obtains original answer to be answered a question.Original answer is according to question and answer relation
Obtaining, the most original answer essence is the dimension from " question-response ", solves the problem that quizmaster proposes.
" asking " in question and answer relation described in the embodiment of the present application refers to the key word that " problem " includes, " answering "
Refer to the knowledge point mark that " answer " is corresponding.In the present embodiment, answer a question according to described waiting and prestore
Question and answer relation, obtain one or more original answer to be answered a question, including: 1) resolve described in treat back
Question and answer is inscribed, and waits, described in acquisition, the predefined key word included of answering a question;2) according to described key word and
Described question and answer relation, obtains the answer with described Keywords matching, as the one or more original answer.
1) wait to answer a question described in parsing, the predefined key word that waiting described in acquisition answers a question includes.
In actual applications, feature to be answered a question is that the key word being contained by characterizes.Therefore basis
Problem obtains answer, substantially obtains answer according to the characteristic key words comprised in problem.Key word is pre-
First define, be stored in keywords database.By keywords database, can treat answers a question carries out participle
Process, find the characteristic key words comprised in problem.Such as, wait to answer a question for: " how to use Alipay
Go to pay?”;In keywords database containing " noun: Alipay ", " noun: remaining sum precious ", " verb: use ",
The content such as " verb: withdraw deposit ", " verb: stolen ", " verb: pay ", then treating answers a question is carried out
After word segmentation processing, the key word that therefrom can extract include " noun: Alipay ", " verb: pay ",
" verb: use ".
2) according to described key word and described question and answer relation, the answer with described Keywords matching is obtained, as
The one or more original answer.
After getting the key word included wait answering a question, according to the question and answer relation between key word and answer,
Retrieval obtains the answer with Keywords matching.Such as: question and answer relation i is: key word " noun: Alipay ",
" verb: pay ", " verb: use " corresponding knowledge point m, the most above-mentioned waiting is answered a question: " how to use
Alipay goes to pay?" original answer include " knowledge point m ".In the present embodiment, by question and answer relation
After retrieving the knowledge point mark that " problem " is corresponding, get the concrete of knowledge point further according to knowledge point mark
Content, particular content can be the information such as the achievement of problem, solution, operating procedure.
After user has browsed some original answer, find that this answer cannot solve problem but this answer is to solve
During the general orientation of certainly problem, it is possible to this answer for retrieval according to performing step S103, so that " result is tied
Relevant dimension really " supplements the single dimension of " problem result ", finds the improvement answer of this original answer,
Thus promote the coverage rate of solution problem.
Step S103: with the information in described undesirable original answer for retrieval foundation, with predetermined
Standard is retrieved in the knowledge point set prestored, with improvement answer to be answered a question described in acquisition;Described know
Know point and include title and content.
The method that the embodiment of the present application provides, is to judge based on such a, it may be assumed that described original answer is not inconsistent
Share the requirement at family, it is impossible to solve problem, but original answer is the general orientation of solution problem.Therefore, logical
During crossing " once result " retrieval " second fruiting ", can be by identifying that original answer and improvement are answered
The feature of case, and the feature of the two is carried out characteristic similarity calculating, user cannot be solved in original answer and ask
The when of topic, recommend an improvement answer more likely solving problem to user.
In actual applications, with the information in described undesirable original answer for retrieval foundation, in advance
The knowledge point set deposited is retrieved, with described in obtaining when the improvement answer answered a question, the retrieval of institute's foundation
Standard can have a variety of specific embodiment, and optional part specific standards is set forth below in the present embodiment.
1) standard one
Standard one refers to the highest with the title similarity of described undesirable original answer.Use this standard,
It is using original answer and to improve the respective title of answer as respective feature, and according to this feature, will be with former
The highest knowledge point of the title similarity of beginning answer is as improvement answer to be answered a question.Use this standard
Theoretical foundation is: owing to the title similarity of original answer and improvement answer is the highest, therefore they have solution
The most like general orientation of problem.Such as: original answer is corresponding " knowledge point i ", improve answer correspondence " to know
Know some j ", then the title similarity of " knowledge point i " and " knowledge point j " is " knowledge point i " and other knowledge
Title similarity between point is the highest.
In actual applications, for improve recall precision, can precalculate and store any two knowledge point it
Between title similarity.When retrieval improves answer, to all title similarities related with original answer
It is ranked up, finds the knowledge point that title similarity peak is corresponding, as improving answer.According to specifically should
By situation, can update arbitrarily to set according to modes such as triggerings behind fixed time interval, renewal knowledge point
Title similarity between two knowledge points.Such as: set the fixing idle time of every day, update any two
Title similarity between individual knowledge point.
Refer to Fig. 3, its be the application intelligent response embodiment of the method in calculate the concrete stream of title similarity
Cheng Tu.In the present embodiment, the mark between any two knowledge point (the first knowledge point and the second knowledge point)
Topic similarity uses following steps to calculate:
Step S301: resolve the first knowledge point and the title of the second knowledge point, obtain described first knowledge point and
The predefined key word that the title of the second knowledge point each includes, respectively as the first title keyword collection
Close and the second title keyword set.
Step S302: according to described first title keyword set and the second title keyword set, calculates institute
State the title similarity of the first knowledge point and the second knowledge point.
According to described first title keyword set and the second title keyword set, calculate the first knowledge point and
The process of the title similarity of the second knowledge point, substantially weighs two set (the first title keyword set
With the second title keyword set) similarity.Jie Kade similarity coefficient is the similarity weighing two set
A kind of index.In the present embodiment, title similarity uses Jie Kade (Jaccard) similarity coefficient.Two
The common factor element of individual set A and B ratio A's and B and shared by concentration, the outstanding card of referred to as two set
Moral similarity coefficient, represents with symbol J (A, B).The title similarity of the first knowledge point and the second knowledge point is permissible
Use following formula to calculate to obtain:
X (A, B)=| A ∩ B |/| A U B |
Wherein, A is described first title keyword set, and B is described second title keyword set, X (A,
B) it is described title similarity.
Owing to the improvement answer the highest with the title similarity of original answer may be multiple, in order to obtain up-to-date,
Most like knowledge point, it is preferred that when the knowledge point the highest with the title similarity of original answer has multiple,
Can select to update the nearest knowledge point of time gap current time as improving answer.
2) standard two
Refer to Fig. 4, its be the application intelligent response embodiment of the method in a concrete stream of step S103
Cheng Tu.With information in described undesirable original answer for retrieval foundation, prestoring with standard two
Knowledge point set is retrieved, with improvement answer to be answered a question described in acquisition, comprises the steps:
Step S401: find out the knowledge point the highest with the title similarity of described undesirable original answer
As candidate knowledge point.
Step S402: judge whether the URL comprised in described candidate knowledge point is not inconsistent with described
Close the original answer required identical.
Improvement answer to be answered a question is obtained, mainly according to standard two retrieval in the knowledge point set prestored
It is divided into two steps: by step S401, finds out the title similarity with described undesirable original answer
High knowledge point is as candidate knowledge point;By step S402, it is judged that the system comprised in described candidate knowledge point
One URLs is the most identical with described undesirable original answer.Use the theoretical foundation of this standard
It is: the general orientation of original answer and improvement answer is the nearest more good, and concrete solution is the most different more good.
The general orientation of original answer and improvement answer is mainly determined by respective title, and concrete solution
Depend on the linked contents of the uniform resource locator that knowledge point includes.If original answer and some knowledge
The title similarity of point is the highest, but content is the most identical, then obviously this knowledge point is not appropriate for as improving answer.
Visible, for standard one, standard two is an embodiment optimized, it is possible to it is the most accurate to find
Improvement project.
Step S403: if above-mentioned judged result is yes, then get rid of this candidate knowledge point, in residue knowledge point
Find out the highest knowledge point of title similarity as described candidate knowledge point, and return previous judgement step.
When the URL comprised in judgement candidate knowledge point is identical with original answer, then by this time
Select knowledge point to get rid of, and according to the order of title similarity, residue knowledge point is found out title similarity
High knowledge point is as new candidate knowledge point, it is judged that the unified resource location comprised in new candidate knowledge point
Accord with the most identical with original answer.Such as: original answer corresponding " knowledge point i ", with its title similarity
High knowledge point is " knowledge point j ", and " knowledge point i " and the uniform resource locator phase of " knowledge point j "
With, it may be assumed that the particular content of " knowledge point i " and " knowledge point j " is identical, then " knowledge point j " arranged
Remove, residue knowledge point finds out the highest knowledge point of title similarity as next candidate knowledge point, continue
Continuous lookup improves answer.
Step S404: if above-mentioned judged result is no, then using described candidate knowledge point as described improvement answer.
When judging that the URL comprised in candidate knowledge point differs with original answer, then illustrate
This candidate knowledge point is different from the solution of original answer, and general orientation is more consistent, can this be known
Know point as improving answer.
3) standard three
Standard three refers to the paragraph heading similarity in the content of described undesirable original answer
High.Use this standard, be as each using the paragraph heading in original answer and the improvement respective content of answer
Feature, and according to this feature, by the knowledge the highest with the paragraph heading similarity in the content of original answer
Point is as improvement answer to be answered a question.
Judge original answer and improve the foundation whether general orientation of answer approximates, both can be the mark of knowledge point
Topic, it is also possible to be the paragraph heading in the content of knowledge point.Therefore, the theoretical foundation of standard three and standard one
Theoretical foundation be similar to: due to original answer and improve answer content in paragraph heading similarity the highest,
Therefore they have the most like general orientation of solution problem.For standard one, standard three only will
Feature is changed to " paragraph heading in content " by " title " of standard one.Standard three and standard one embodiment
The part that content is identical repeats no more, and refers to the appropriate section in standard one embodiment.
In actual applications, for improve recall precision, can precalculate and store any two knowledge point it
Between content in paragraph heading similarity.When retrieval improves answer, to institute related with original answer
There is the paragraph heading similarity in content to be ranked up, find the paragraph heading similarity peak pair in content
The knowledge point answered, as improving answer.According to concrete applicable cases, between can setting according to regular time
Every, update behind knowledge point the modes such as triggering, update the paragraph heading in the content between any two knowledge point
Similarity.Such as: set the fixing idle time of every day, update in the content between any two knowledge point
Paragraph heading similarity.
In the present embodiment, the content between any two knowledge point (the first knowledge point and the second knowledge point)
In paragraph heading similarity use following steps calculate:
Step S501: resolve the described paragraph heading of the first knowledge point and the second knowledge point, obtain described first
The predefined key word that the described paragraph heading of knowledge point and the second knowledge point each includes, by described
The key word that the described paragraph heading of one knowledge point includes is as the keyword set of the first paragraph heading, by institute
State key word that the described paragraph heading of the second knowledge point the includes keyword set as the second paragraph heading.
Step S502: according to keyword set and the key word of the second paragraph heading of described first paragraph heading
Set, calculates the paragraph heading similarity in the content of described first knowledge point and the second knowledge point.
Identical with standard one, according to keyword set and the pass of the second paragraph heading of described first paragraph heading
Keyword set, calculates the process of paragraph heading similarity in the content of the first knowledge point and the second knowledge point,
Substantially weigh two set (keyword set of the first paragraph heading and the keyword set of the second paragraph heading
Close) similarity.In the present embodiment, the paragraph heading similarity in content uses Jie Kade similarity coefficient.
Paragraph heading similarity in the content of the first knowledge point and the second knowledge point can use following formula to calculate and obtain
Take:
Y (C, D)=| C ∩ D |/| C U D |
Wherein, C is the keyword set of described first paragraph heading, and D is the key of described second paragraph heading
Set of words, Y (C, D) is the paragraph heading similarity in described content.
Owing to the improvement answer the highest with the paragraph heading similarity in the content of original answer may be multiple,
In order to obtain up-to-date, most like knowledge point, it is preferred that when with the paragraph heading in the content of original answer
When the knowledge point that similarity is the highest has multiple, select to update the knowledge point conduct that time gap current time is nearest
Improve answer.
4) standard four
With the information in described undesirable original answer for retrieval foundation, with standard four at knowing of prestoring
Know in some set and retrieve, with improvement answer to be answered a question described in acquisition, comprise the steps:
Step S601: find out and the paragraph heading similarity in the content of described undesirable original answer
The highest knowledge point is as candidate knowledge point.
Step S602: judge whether the URL comprised in described candidate knowledge point is not inconsistent with described
Close the original answer required identical;The most then enter step S603;If it is not, then enter step S604.
Similar with standard two, obtain to be answered a question according to standard four retrieval in the knowledge point set prestored
Improve answer, be broadly divided into two steps: by step S601, find out and described undesirable original answer
Content in the highest knowledge point of paragraph heading similarity as candidate knowledge point;By step S602, sentence
The URL comprised in disconnected described candidate knowledge point whether with described undesirable original answer
Identical.For standard two, feature is only changed to " interior by " title " of standard two by standard four
Paragraph heading in appearance ", standard four is original answer and the paragraph heading improved in the respective content of answer to be made
For respective feature, and according to this feature, by the highest with the paragraph heading similarity in the content of original answer
Knowledge point as candidate knowledge point.The part that standard four is identical with standard two embodiment content repeats no more,
Refer to the appropriate section in standard two embodiment.
The theoretical foundation using this standard is: original answer and to improve the general orientation of answer the nearest more good, and has
The solution of body is the most different more good.In standard four, the general orientation of original answer and improvement answer is mainly
Determined by the paragraph heading in respective content, and concrete solution depends on the system that knowledge point includes
The linked contents of one resource localizer.If original answer and the paragraph heading in the content of some knowledge point
Similarity is the highest, and content is the most identical, then obviously this knowledge point is not appropriate for as improving answer.It is visible,
For standard three, standard four is an embodiment optimized, it is possible to finds and improves the most accurately
Scheme.
Step S603: if above-mentioned judged result is yes, then get rid of this candidate knowledge point, in residue knowledge point
Find out the highest knowledge point of the paragraph heading similarity in content as described candidate knowledge point, and return previous
Judge step.
Step S604: if above-mentioned judged result is no, then using described candidate knowledge point as described improvement answer.
5) standard five
Standard five be according to the paragraph in the title similarity of described undesirable original answer, content
Title similarity and uniform resource locator similarity, calculate overall similarity, with overall similarity with described
The undesirable immediate knowledge point of original answer is as improving knowledge point;Described overall similarity uses
Formula calculated as below:
W=X*N1+Y*N2 Z*N3
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator.
Standard five considers the paragraph heading similarity in title similarity, content and uniform resource locator
Similarity, calculates the overall similarity obtaining two knowledge points, therefore relative to aforementioned four standard, standard
Five is a preferred standard.The theoretical foundation using this standard is: original answer and improve the generous of answer
To the nearest more good, and concrete solution is the most different more good.In standard five, the judgement of general orientation considers
Paragraph heading two parts in knowledge point title and content, thus the judged result of general orientation is the most accurate.
In the present embodiment, calculated by lot of experiments that to obtain the weight of title similarity, paragraph heading similar
The weight of degree and the empirical value of the weight of uniform resource locator similarity, be respectively as follows: the power of title similarity
Weight values N1 is 100, the weighted value of paragraph heading similarity is 10, the weight of uniform resource locator similarity
Value is 100.Each weighted value above-mentioned is an empirical value obtained by lot of experiments, and weighted value can also lead to
The mode crossing machine learning obtains automatically.Visible, in the embodiment of the present application by the empirical value of each weight
The weighted value of paragraph heading similarity is significantly less than the weighted value of title similarity, it is seen that relative to paragraph heading
For similarity, title similarity is the principal element determining general orientation, and this is consistent with practical situation.
Further, as it is desirable that solution concrete between two knowledge points is the most different more good, therefore by paragraph heading
Similarity and title similarity sum, deduct uniform resource locator similarity, calculates and obtains overall similarity.
Wherein, the portion that title similarity is identical with above-mentioned standard content with the computational methods of paragraph heading similarity
Divide and repeat no more, refer to the appropriate section in relevant criterion.Any two knowledge point (the first knowledge point and
Second knowledge point) between content in uniform resource locator similarity use following steps calculate:
Step S701: resolve each content of the first knowledge point and the second knowledge point, obtains described first knowledge
Point and the uniform resource locator each included of the second knowledge point, respectively as the first uniform resource locator
Set and the second uniform resource locator set.
Step S702: according to described first uniform resource locator set and the second uniform resource locator set,
Calculate the described uniform resource locator similarity of described first knowledge point and the second knowledge point.
In the present embodiment, the uniform resource locator similarity in content uses Jie Kade similarity coefficient.The
Uniform resource locator similarity in the content of one knowledge point and the second knowledge point can use following formula meter
Calculate and obtain:
Z (E, F)=| E ∩ F |/| E U F |
Wherein, E is described first uniform resource locator set, and F is described second uniform resource locator collection
Closing, Z (E, F) is described uniform resource locator similarity.
6) standard six
Standard six is an optimisation criteria of standard five, and this standard also contemplates the renewal time factor of knowledge point,
For the knowledge point in the default renewal time, according to the renewal time that it is concrete, calculate overall similarity.More
The knowledge point that the new time is the nearest, its overall similarity is the highest.Employing standard six obtains improves answer, including such as
Lower step:
Step S801: judge whether current time is less than with the time interval of the recent renewal time of knowledge point
In the first prefixed time interval.
Step S802: if above-mentioned judged result is yes, then basis and described undesirable original answer
Paragraph heading similarity in title similarity, content and uniform resource locator similarity, and described work as
Front time and the time interval of the recent renewal time of described knowledge point, calculate overall similarity, with described whole
Body similarity and the described undesirable immediate knowledge point of original answer are as improving knowledge point;Described
Overall similarity employing formula calculated as below:
W=X*N1+Y*N2 Z*N3+N4* (Ts-t)
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator, when N4 is default
Between weight, Ts is described first prefixed time interval, t be described current time with described knowledge point the most more
The time interval of new time.
The part that standard six is identical with standard five content repeats no more, and refers to the appropriate section in standard five.
In the formula of standard six, N4 is default time weighting, and Ts is described first prefixed time interval, and these are two years old
Individual weighted value also can be by the empirical value that lot of experiments obtains, it is also possible to by the way of machine learning certainly
Dynamic acquisition.Such as: N4 value is 3, Ts value is 10, t is 5, and its implication is: two knowledge points it
Between overall similarity on the basis of standard five, add 15 points, it may be assumed that update time the nearest knowledge point,
Its overall similarity is the highest.
With standard one to standard four in like manner, for standard five and standard six, in actual applications, for carrying
High recall precision, can precalculate and store the overall similarity between any two knowledge point.In retrieval
When improving answer, all overall similarities related with original answer are ranked up, find global similarity
The knowledge point that degree peak is corresponding, as improving answer.According to concrete applicable cases, can set according to solid
Fixed time interval, update behind knowledge point the modes such as triggering, update the overall phase between any two knowledge point
Like degree.Such as: set the fixing idle time of every day, the global similarity between any two knowledge point is updated
Degree.
By above-mentioned six kinds of standards, with the information in described undesirable original answer for retrieval foundation,
Retrieval the improvement answer of different order of accuarcys to be answered a question can be obtained in the knowledge point set prestored.
In actual applications, suitable standard can be selected according to real needs, obtain the improvement answer of original answer.
When implementing the method that the application provides, it is contemplated that the new and old factor of knowledge point, various retrievals are changed
Enter the preassigned of answer, it is also possible to include on the basis of each standard: only by stipulated time threshold range
The knowledge point of interior renewal obtains the scope of described improvement answer as retrieval;Described time threshold is current time
Preset value with the time interval length of recent renewal time of described knowledge point.With it, can limit
Surely the renewal time of answer is improved, it is to avoid knowledge point earlier, such as: various station addresses are the most expired,
Or the problems such as the software correcting of the problem of solution, thus get up-to-date knowledge point.Such as: by time
Between threshold value be set as 365 days, then the program obtain improve answer the renewal time one be set to 365 days in more
New knowledge point.
In the above-described embodiment, it is provided that a kind of intelligent response method, corresponding, the application is also
A kind of intelligent response device is provided.Refer to Fig. 5, it is the signal of intelligent response device embodiment of the application
Figure.This device is corresponding with the embodiment of above-mentioned intelligent response method.
Owing to device embodiment is substantially similar to embodiment of the method, so describing fairly simple, relevant part
The part seeing embodiment of the method illustrates.Device embodiment described below is only schematically.
A kind of intelligent response device of the present embodiment, including acquiring unit 101, is used for obtaining corresponding specific treat
The person of being asked answered a question is evaluated as undesirable original answer;Retrieval unit 103, for described
Information in undesirable original answer is retrieval foundation, with predetermined standard at the knowledge point set prestored
Conjunction is retrieved, with improvement answer to be answered a question described in acquisition;Described knowledge point includes title and content.
Refer to Fig. 6, its be the application intelligent response device embodiment in the concrete signal of retrieval unit 103
Figure.Optionally, described retrieval unit 103 includes:
First retrieval subelement 1031, for the information in described undesirable original answer for retrieval
Foundation, is up to retrieval with the title similarity of described improvement answer Yu described undesirable original answer
Standard, retrieves in the described knowledge point prestored is gathered, and answers with described improvement to be answered a question described in acquisition
Case.
Optionally, described retrieval unit 103 includes:
Second retrieval subelement 1032, for the information in described undesirable original answer for retrieval
Foundation, retrieves in the described knowledge point prestored is gathered, and answers with described improvement to be answered a question described in acquisition
Case.
Refer to Fig. 7, its be the application intelligent response device embodiment in the tool of the second retrieval unit 1032
Body schematic diagram.Optionally, described second retrieval subelement 1032 includes:
First searches subelement 10321, similar to the title of described undesirable original answer for finding out
Spend the highest knowledge point as candidate knowledge point;
Judgment sub-unit 10322, for judging the URL that comprises in described candidate knowledge point whether
Identical with described undesirable original answer;
Second searches subelement 10323, if the URL comprised in the described candidate knowledge point with
Described undesirable original answer is identical, then get rid of this candidate knowledge point, finds out in residue knowledge point
The highest knowledge point of title similarity is as described candidate knowledge point, and returns previous judgement step;
Judge subelement 10324, if the URL comprised in described candidate knowledge point is with described
Undesirable original answer differs, then using described candidate knowledge point as described improvement answer.
Optionally, described retrieval unit 103 includes:
3rd retrieval subelement 1033, for the information in described undesirable original answer for retrieval
Foundation, similar to the paragraph heading in the content of described undesirable original answer with described improvement answer
Degree is up to search criteria, retrieves in the described knowledge point prestored is gathered, and waits to answer a question described in obtaining
Described improvement answer.
Optionally, described retrieval unit 103 includes:
4th retrieval subelement 1034, for the information in described undesirable original answer for retrieval
Foundation, retrieves in the described knowledge point prestored is gathered, and answers with described improvement to be answered a question described in acquisition
Case;
Described 4th retrieval subelement 1034 includes:
First search subelement 10341, for find out with in the content of described undesirable original answer
The highest knowledge point of paragraph heading similarity is as candidate knowledge point;
Judgment sub-unit 10342, for judging the URL that comprises in described candidate knowledge point whether
Identical with described undesirable original answer;
Second searches subelement 10343, if the URL comprised in the described candidate knowledge point with
Described undesirable original answer is identical, then get rid of this candidate knowledge point, finds out in residue knowledge point
The highest knowledge point of paragraph heading similarity in content is as described candidate knowledge point, and returns previous judgement
Step;
Judge subelement 10344, if the URL comprised in described candidate knowledge point is with described
Undesirable original answer differs, then using described candidate knowledge point as described improvement answer.
Optionally, described retrieval unit 103 includes:
5th retrieval subelement 1035, for the information in described undesirable original answer for retrieval
Foundation, is up to retrieval with the overall similarity of original improvement answer Yu described undesirable original answer
Standard, retrieves in the described knowledge point prestored is gathered, and answers with described improvement to be answered a question described in acquisition
Case;
Described overall similarity be according to the title similarity of described undesirable original answer, content
In paragraph heading similarity and uniform resource locator Similarity Measure obtain;Described overall similarity is adopted
Use formula calculated as below:
W=X*N1+Y*N2 Z*N3
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator.
Optionally, described retrieval unit 103 includes:
6th retrieval subelement 1036, for the information in described undesirable original answer for retrieval
Foundation, is up to retrieval with the overall similarity of described improvement answer Yu described undesirable original answer
Standard, retrieves in the described knowledge point prestored is gathered, and answers with described improvement to be answered a question described in acquisition
Case;
Described overall similarity be according to the title similarity of described undesirable original answer, content
In paragraph heading similarity and uniform resource locator similarity, and described current time and knowledge point
The time interval of recent renewal time calculates acquisition;Described overall similarity employing formula calculated as below:
W=X*N1+Y*N2 Z*N3+N4* (Ts-t)
Wherein, W is described overall similarity, and X is described title similarity, and Y is described paragraph heading phase
Like degree, Z is described uniform resource locator similarity, and N1 is the weight of described title similarity, and N2 is institute
Stating the weight of paragraph heading similarity, N3 is the weight of described uniform resource locator, when N4 is default
Between weight, Ts is described default time interval, and t is the recent renewal of described current time and described knowledge point
The time interval of time.
Refer to Fig. 8, it is the concrete schematic diagram of intelligent response device embodiment of the application.Optionally, its
Being characterised by, described device also includes:
First computation subunit 201, for calculating the described title similarity of any two knowledge point.
Refer to Fig. 9, its be the application intelligent response device embodiment in the first computing unit 201 concrete
Schematic diagram.Optionally, described first computation subunit 201 includes:
Resolve subelement 2011, for resolving the first knowledge point and the title of the second knowledge point, obtain described the
The predefined key word that the title of one knowledge point and the second knowledge point each includes, respectively as the first mark
Topic keyword set and the second title keyword set;
Computation subunit 2012, for according to described first title keyword set and the second title keyword collection
Close, calculate the title similarity of described first knowledge point and the second knowledge point.
Optionally, described device also includes:
Second computation subunit 202, the paragraph heading phase in the described content calculating any two knowledge point
Like degree;
Described second computation subunit 202 includes:
Resolve subelement 2021, for resolving the first knowledge point and the described paragraph heading of the second knowledge point, obtain
Take the predefined key that the described paragraph heading of described first knowledge point and the second knowledge point each includes
Word, the key word included by the described paragraph heading of described first knowledge point is as the key of the first paragraph heading
Set of words, the key word included by the described paragraph heading of described second knowledge point is as the second paragraph heading
Keyword set;
Computation subunit 10382, for the keyword set according to described first paragraph heading and the second paragraph mark
The keyword set of topic, calculates described first knowledge point similar with the paragraph heading in the content of the second knowledge point
Degree.
Optionally, described device also includes:
3rd computation subunit 203, the unified resource in the described content calculating any two knowledge point is fixed
Position device similarity;
Described 3rd computation subunit 203 includes:
Resolve subelement 2031, for resolving the first knowledge point and each content of the second knowledge point, obtain institute
State the uniform resource locator each included of the first knowledge point and the second knowledge point, unified respectively as first
Resource localizer set and the second uniform resource locator set;
Computation subunit 2032, for according to described first uniform resource locator set and the second unified resource
Localizer set, calculates the described uniform resource locator similarity of described first knowledge point and the second knowledge point.
Optionally, described device also includes:
4th computation subunit 204, for calculating the described overall similarity of any two knowledge point.
Optionally, described device also includes:
Signal generating unit 205, is used for generating described undesirable original answer.
Refer to Figure 10, its be the application intelligent response device embodiment in specifically the showing of signal generating unit 205
It is intended to.Optionally, described signal generating unit 205 includes:
Receive subelement 2051, for receiving inquiry request to be answered a question described in the correspondence that client sends;
Obtain subelement 2052, for the question and answer relation waiting to answer a question and prestore described in basis, obtain described
The original answer of to be answered a question the one or more;
Choose subelement 2053, for one quizmaster chosen from the one or more original answer
Original answer, as described undesirable original answer.
Optionally, described acquisition subelement 2052 includes:
Resolve subelement 20521, be used for resolving described in wait to answer a question, waiting described in acquisition answers a question includes
Predefined key word;
Inquiry subelement 20522, for according to described key word and described question and answer relation, obtains and described key
The answer of word coupling, as the one or more original answer.
The embodiment of the present application additionally provides a kind of intelligent response system, including above-mentioned intelligent response device.
Additionally, the embodiment of the present application additionally provides a kind of electronic equipment, as shown in figure 11, this electronic equipment bag
Include: display 1101;Processor 1102;And memorizer 1103, described memorizer 1103 is configured to
Store and wait to answer a question, described in when being performed by described processor 1102 wait answering a question, at described display 1101
Improvement answer to be answered a question described in display, described improvement answer is that the person of being asked is evaluated as not meeting and wants
Information in the original answer asked is retrieval foundation, retrieves in the knowledge point set prestored with predetermined standard
Obtaining, described original answer is to wait described in basis that the question and answer relation answered a question and prestore generates.
Electronic equipment described in the embodiment of the present application includes the terminal unit such as PC, PAD, iPad, and
Mobile communication equipment, it may be assumed that usually said mobile phone or smart mobile phone.Electronic equipment can also is that Wearable
Smart machine, such as, Fructus Mali pumilae wrist-watch or Google's glasses.
Knowledge point and question and answer relation described in the embodiment of the present application both can store in the electronic device, it is also possible to
It is stored in the server networked with electronic equipment.Knowledge point and question and answer relation storage in the electronic device excellent
Point is, the speed answered a question is faster;And shortcoming is to be managed collectively knowledge point and question and answer relation.
In the present embodiment, electronic equipment needs and mutual as a networked devices, the enquirement of problem and answer
Server interaction in networking.Therefore, described knowledge point and described question and answer relation are stored in server end, and
Non-memory is in the memorizer of electronic equipment.
By knowledge point and question and answer relation are stored in server end, it is possible to unified management knowledge point and question and answer are closed
System, only needs when updating knowledge point and question and answer relation to update server end, it is to avoid by knowledge point and question and answer relation
Store the problem that the data that may cause in the electronic device are inconsistent.
In actual applications, after electronic equipment gets the improvement answer of problem, it is also possible to answer will be improved
It is stored in the memorizer of electronic equipment, in order to follow-up when electronic equipment off-line, it is possible to have obtained at any time
The improvement answer of the problem through answering.
Intelligent response method, device, system and the electronic equipment that the application provides, corresponding specific by obtaining
The to be answered a question person of being asked is evaluated as undesirable original answer, and with described undesirable
Information in original answer is retrieval foundation, retrieves in the knowledge point set prestored with predetermined standard, with
Improvement answer to be answered a question described in acquisition so that query and search to wait to answer a question the most corresponding
After answer, additionally it is possible to carry out quadratic search according to the undesirable original answer that quizmaster selects, obtain
Improvement answer to be answered a question.Owing to quizmaster reflects quizmaster to answer to the selection of original answer
Demand, can be as new retrieval foundation.Therefore, this method can be difficult to lead in prior art
In the case of crossing the information raising answer coverage rate and accuracy rate that waiting answers a question itself comprises, new by introducing
Retrieval according to improving the coverage rate of answer and accuracy rate.
Although the application is open as above with preferred embodiment, but it is not for limiting the application, Ren Heben
Skilled person, without departing from spirit and scope, can make possible variation and amendment,
Therefore the protection domain of the application should be defined in the range of standard with the application claim.
In a typical configuration, calculating equipment includes one or more processor (CPU), input/output
Interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
(RAM) and/or the form such as Nonvolatile memory, such as read only memory (ROM) or flash memory (flash RAM).
Internal memory is the example of computer-readable medium.
1, computer-readable medium includes that permanent and non-permanent, removable and non-removable media can be by
Any method or technology realize information storage.Information can be computer-readable instruction, data structure, journey
The module of sequence or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory
(PRAM), static RAM (SRAM), dynamic random access memory (DRAM), its
The random access memory (RAM) of his type, read only memory (ROM), electrically erasable is read-only deposits
Reservoir (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read only memory
(CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tape, tape magnetic magnetic
Disk storage or other magnetic storage apparatus or any other non-transmission medium, can be used for storage can be set by calculating
The standby information accessed.According to defining herein, computer-readable medium does not include non-temporary computer-readable matchmaker
Body (transitory media), such as data signal and the carrier wave of modulation.
2, it will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer
Program product.Therefore, the application can use complete hardware embodiment, complete software implementation or combine software
Form with the embodiment of hardware aspect.And, the application can use and wherein include meter one or more
The computer-usable storage medium of calculation machine usable program code (include but not limited to disk memory,
CD-ROM, optical memory etc.) form of the upper computer program implemented.