CN105912697A - Optimization method and device of dialog system knowledge base - Google Patents
Optimization method and device of dialog system knowledge base Download PDFInfo
- Publication number
- CN105912697A CN105912697A CN201610262475.7A CN201610262475A CN105912697A CN 105912697 A CN105912697 A CN 105912697A CN 201610262475 A CN201610262475 A CN 201610262475A CN 105912697 A CN105912697 A CN 105912697A
- Authority
- CN
- China
- Prior art keywords
- information
- answer
- answer information
- question
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an optimization method and device of a dialog system knowledge base. The optimization method comprises the following steps: obtaining target question and answer information: judging whether assessment parameters of question and answer information to be analyzed conform to an optimization standard or not, and taking the question and answer information to be analyzed as the target question and answer information when the assessment parameters of the question and answer information to be analyzed conform to the optimization standard; and generating an update answer set: on the basis of question information of the target question and answer information, obtaining a corresponding answer information set, independently calculating a relevance parameter of each piece of answer information in the answer information set, and generating the update answer set which aims at the question information in the target question and answer information according to the relevance parameters. The method can realize automatic update of the dialog system knowledge base, so that a dialog system can output answers which more conform to the use habit and the expectation of users so as to improve the user experience and the user viscosity of the dialog system.
Description
Technical field
The present invention relates to human-computer interaction technique field, specifically, relate to the optimization of a kind of conversational system knowledge base
Method and device.
Background technology
For traditional man-machine interaction, man-machine interaction is mainly passed through mouse, keyboard and touch by user
The devices such as screen interact with the equipment such as computer, mobile phone.And along with between man-machine information interaction amount just present
Volatile growth, traditional man-machine interaction mode greatly have impact on efficiency and the effect of man-machine interaction.
People be accustomed to the most natural, the most easily interactive mode be that natural language is mutual, therefore by dialogue be
System realizes man-machine interaction efficiently becomes the most universal selection.But, existing conversational system for
Some inquiry problems of user cannot provide gratifying answer, and this most just have impact on the use of whole conversational system
Family is experienced so that user's viscosity of product is poor.
Summary of the invention
For solving the problems referred to above, the invention provides the optimization method of a kind of conversational system knowledge base, including:
Target question and answer information acquiring step, it is judged that whether the test and appraisal parameter of question and answer information to be analyzed meets optimization mark
Standard, as met, then using described question and answer information to be analyzed as target question and answer information;
Updating answer set generation step, problem information based on described target question and answer information obtains corresponding answer
Information aggregate, calculates the degree of association parameter of each answer information in described answer information set respectively, and according to institute
State degree of association parameter to generate for the renewal answer set of problem information in described target question and answer information.
According to one embodiment of present invention, in described target question and answer information acquiring step, it is judged that described in treat point
Whether the test and appraisal parameter of analysis question and answer information is less than is preset test and appraisal threshold value, if it is less, by described question and answer to be analyzed
Information is as target question and answer information.
According to one embodiment of present invention, in described renewal answer set generation step, according to question and answer information
Whether middle problem information has the identical centre word of semanteme with answer information is joined to the degree of association determining answer information
Number, wherein, the quantity of semantic identical centre word is the most, and the degree of association parameter of answer information and problem information is more
Greatly.
According to one embodiment of present invention, when obtaining described answer information set, choose according to preset rules
Targeted customer, and push the problem information in described target question and answer information to described targeted customer, obtain described mesh
The answer information that mark user is fed back for described problem information, thus obtain described question and answer information aggregate.
According to one embodiment of present invention, when choosing described targeted customer,
Obtain the user property of conversational system different user, it is judged that whether user property meets the propelling movement of default problem is wanted
Asking, if met, then corresponding user being defined as described targeted customer, wherein, described user property includes
Any one in item set forth below or several:
Subscriber identity information, customer position information, age of user information, user gender information;
Or, obtain the interaction scenarios of conversational system different user and/or mutual topic, it is judged that described interaction scenarios
And/or whether topic meets the propelling movement requirement of default problem alternately, if met, then corresponding user is defined as institute
State targeted customer.
Present invention also offers the optimization device of a kind of conversational system knowledge base, including:
Target question and answer data obtaining module, it is for judging whether the test and appraisal parameter of question and answer information to be analyzed meets excellent
Change standard, as met, then using described question and answer information to be analyzed as target question and answer information;
Updating answer set generation module, it obtains correspondence for problem information based on described target question and answer information
Answer information set, calculate the degree of association parameter of each answer information in described answer information set respectively, and
Generate for the renewal answer set of problem information in described target question and answer information according to described degree of association parameter.
According to one embodiment of present invention, described target question and answer data obtaining module be configured to judge described in treat point
Whether the test and appraisal parameter of analysis question and answer information is less than is preset test and appraisal threshold value, if it is less, by described question and answer to be analyzed
Information is as target question and answer information.
According to one embodiment of present invention, described renewal answer set generation module is configured to according to question and answer information
Whether middle problem information has the identical centre word of semanteme with answer information is joined to the degree of association determining answer information
Number, wherein, the quantity of semantic identical centre word is the most, and the degree of association parameter of answer information and problem information is more
Greatly.
According to one embodiment of present invention, described renewal answer set generation module is configured to described in obtaining answer
During case information aggregate, choose targeted customer according to preset rules, and push described target to described targeted customer and ask
Answer the problem information in information, obtain the answer information that described targeted customer is fed back for described problem information,
Thus obtain described question and answer information aggregate.
According to one embodiment of present invention, described renewal answer set generation module is configured to choosing described mesh
During mark user,
Obtain the user property of conversational system different user, it is judged that whether user property meets the propelling movement of default problem is wanted
Asking, if met, then corresponding user being defined as described targeted customer, wherein, described user property includes
Any one in item set forth below or several:
Subscriber identity information, customer position information, age of user information, user gender information;
Or, obtain the interaction scenarios of conversational system different user and/or mutual topic, it is judged that described interaction scenarios
And/or whether topic meets the propelling movement requirement of default problem alternately, if met, then corresponding user is defined as institute
State targeted customer.
Conversational system knowledge base optimization method provided by the present invention is by the user's Request Log to mass users
It is analyzed, finds bad in these daily records or irrational question and answer information as far as possible, then by various canals
Road obtains the answer information of the problem information in this kind of question and answer information on one's own initiative, and answer information is carried out legitimacy
Judgement, finally by inserted or updated for satisfactory answer information in data base.By to conversational system number
According to the renewal in storehouse, conversational system can export and more conform to user's use habit and desired answer, thus carries
The Consumer's Experience of high conversational system and user's viscosity.
The method achieve self study and the closed loop feedback of conversational system knowledge base.Wherein, the method can make
Bad in knowledge base and irrational question and answer information independence can be carried out rationalizing and improves by conversational system,
So that the quality of this system can have the lifting of persistence.Meanwhile, the method can also make conversational system
Realize autonomic learning, the most just can depart from the manual learning style having to rely on attendant.
The method by the problem information in target question and answer information is pushed to again user obtain user for
The answer of this problem information, the most just achieves the closed loop feedback of " user-conversational system-user ", the most right
Telephone system the most just can keep the stability of self when by external interference.It is based on this closed-loop structure,
What conversational system can continue receives the feedback from user's perception aspect, and comes not according to the feedback of user
Disconnected optimize its knowledge storehouse thus adjust the output of self, so that this output can meet the expectation of user.
Additionally, conversational system knowledge base optimization method provided by the present invention can also be by the attribute according to user
The problem information in target question and answer information is pushed to specific user.Due to the attribute of this kind of user and to dialogue
In system input target question and answer information, the attribute of the user of problem information is same or like, therefore by this kind of user
Also it is obtained with being directed to the higher answer of the most accurate or compatible degree of problem information in target question and answer information
Information, and the renewal answer set of the problems referred to above information determined according to this kind of answer information also will more be as the criterion
Really.
Other features and advantages of the present invention will illustrate in the following description, and, partly from description
In become apparent, or by implement the present invention and understand.The purpose of the present invention and other advantages can be passed through
Structure specifically noted in description, claims and accompanying drawing realizes and obtains.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment
Or the accompanying drawing required in description of the prior art does and simply introduces:
Fig. 1 is the flowchart of conversational system knowledge base optimization method according to an embodiment of the invention;
Fig. 2 is the flowchart of conversational system knowledge base optimization method in accordance with another embodiment of the present invention;
Fig. 3 is the flowchart of the conversational system knowledge base optimization method according to another embodiment of the present invention;
Fig. 4 is the structural representation that conversational system knowledge base optimizes device according to an embodiment of the invention.
Detailed description of the invention
Embodiments of the present invention are described in detail, whereby to the present invention such as below with reference to drawings and Examples
What application technology means solves technical problem, and the process that realizes reaching technique effect can fully understand and evidence
To implement.As long as it should be noted that do not constitute conflict, each embodiment in the present invention and respectively implementing
Each feature in example can be combined with each other, the technical scheme formed all protection scope of the present invention it
In.
Meanwhile, in the following description, many details are elaborated for illustrative purposes, to provide this
The thorough understanding of inventive embodiments.It will be apparent, however, to one skilled in the art, that this
Bright can detail here or described ad hoc fashion implement.
It addition, can be at the meter of such as one group of computer executable instructions in the step shown in the flow chart of accompanying drawing
Calculation machine system performs, and, although show logical order in flow charts, but in some situation
Under, can be to be different from the step shown or described by order execution herein.
For the conversational system in intelligent robot, from the perspective of user, the best Consumer's Experience,
It means that the highest user's viscosity.For dialogue robot, experience the most intuitively is
Whether the answer of the problem that user is inquired meets the expectation of user.Such as, when user is to dialogue machine Crinis Carbonisatus
When playing inquiry " the most alone ", for existing dialogue robot, the answer of its feedback is likely to
It is " I don't have the slightest idea about the meaning of your remarks ".It can thus be seen that existing dialogue robot cannot expire well
The daily interaction demand of foot user.
Dialogue robot is to utilize conversational system mutual to the dialogue realizing between robot and user, wherein,
Conversational system utilizes the knowledge base of self to determine corresponding answer according to the problem that user inputs.The present invention
Utilize above-mentioned characteristic just, it is provided that the optimization method of a kind of new conversational system knowledge base, so that dialogue system
Knowledge base after system can utilize optimization more accurately, reasonably determines the answer of problem that user inputted.
This optimization method, by continuous learning incessantly, constantly promotes the quality of conversational system knowledge base with this.
It is pointed out that the lifting of the quality of the conversational system knowledge base of indication of the present invention, both can refer to knowledge
The improvement of answer information corresponding to problem information in storehouse, it is also possible to refer to problem information and answer letter thereof in knowledge base
The expansion of breath.
In order to clearly illustrate conversational system knowledge base optimization method provided by the present invention realize principle,
Realize process and advantage, below in conjunction with different embodiments, the method is further described.
Embodiment one:
Fig. 1 shows the flowchart of the optimization method of the conversational system knowledge base that the present embodiment provided.
As it is shown in figure 1, first the method that the present embodiment is provided obtains question and answer to be analyzed letter in step S101
The test and appraisal parameter of breath.Wherein, the test and appraisal parameter that the method is accessed in step S101 is preferably question and answer
The scoring to question and answer information to be analyzed of the evaluation system.For the problem information in question and answer information, its answer is believed
The scoring of breath is the highest, and also just explanation user is the highest to the satisfaction of this answer information;Whereas if answer letter
The scoring of breath is the lowest, and also just explanation user is the lowest to the satisfaction of this answer information, and this kind of answer information is also
Need the information being optimized just.
Therefore, in the present embodiment, the method judges the test and appraisal obtained in step S101 in step s 102
Whether parameter meets is preset optimisation criteria.If test and appraisal parameter meets presets optimisation criteria, the party's rule is in step
Using the question and answer information to be analyzed in step S101 as target question and answer information in S103.Wherein, target question and answer letter
Breath is question and answer information to be optimized.
Specifically, in the present embodiment, the test and appraisal parameter acquired in step S101 due to the method is question and answer
Evaluation system is for the scoring of answer information, and therefore the method is in step s 102 by judging that test and appraisal parameter is
No less than preset test and appraisal threshold value judge whether question and answer information to be analyzed is target question and answer information.
Such as the problem information " the most alone " of user's input, what conversational system was fed back answers
Case information " I don't have the slightest idea about the meaning of your remarks " obviously cannot make user please oneself, the therefore scoring of its correspondence
The most just will necessarily be the lowest.If the scoring corresponding to this answer information is less than presetting scoring threshold value, then just
Represent need this answer information is optimized so that user input problem information time conversational system can be anti-
Feedback makes the answer that user is more satisfied with.
As it is shown in figure 1, after determining target question and answer information, the method is asked based on target in step S104
The problem information answering information obtains corresponding answer information aggregate.In the present embodiment, when determining conversational system number
After the target question and answer information in storehouse, the method can be believed the problem in target question and answer information under specific scene
Breath is pushed to the user of conversational system, and records user's feedback to this problem information, thus obtains problem information
Corresponding answer information aggregate.
Such as, the method in step s 103 determined by target question and answer information include: problem information " total from
Oneself is a people " and answer information " I don't have the slightest idea about the meaning of your remarks ", in step S104, the method meeting
Problem information " the most alone " is pushed to conversational system user (user herein may for input should
The user of problem information is likely to as other multiple different users).For the problem information pushed, user
Can be by the answer feedback of oneself to conversational system in dialog procedure, such the method is by collecting the feedback of user
Answer just can obtain the answer information set in target question and answer information corresponding to problem information.
It is pointed out that in other embodiments of the invention, the method can also pass through other rational methods
Obtain corresponding to the answer information set of problem information in target question and answer information, the invention is not restricted to this.
After obtaining answer information set, it is mostly by multiple owing to answering the answer information included in information aggregate
Different users is fed back, and the most both may comprise the answer information that user can be made more to be satisfied with, also
May comprise and cannot make customer satisfaction system answer information, the most also be accomplished by answer information aggregate is screened,
To be determined to the information of the alternative answer of problem information in target question and answer information.
Specifically, as it is shown in figure 1, after obtaining answer information set, the method is difference in step S105
Calculate the degree of association parameter of each answer information in answer information set, and in step s 106 according to degree of association
Parameter generates for the renewal answer information set of problem information in target question and answer information.
In the present embodiment, degree of association parameter is used for characterizing the degree of contact between answer information and problem information, and
An always important research class in artificial intelligence field is contacted between decision problem information and answer information
Topic.So that conversational system can work as efficiently as possible, in the present embodiment, the method utilizes and presets
Rule calculate the degree of association parameter of each answer information in answer information set.
Specifically, in the present embodiment, the method utilizes multiple rule being applicable to degree of association judgement to calculate respectively
The degree of association of question and answer information under each rule, followed by each rule weight to utilize each rule calculated
The degree of association obtained is weighted, thus finally gives the degree of association parameter that a certain answer information is overall.
Wherein, the preset rules that the method is utilized preferably includes centre word rule, and the method is believed according to question and answer
In breath, whether answer information has the identical centre word of semanteme to determine the degree of association of answer information with problem information
Parameter.If the centre word that problem information is identical with semanteme in answer information is the most, then answer information and problem
The value of the degree of association parameter of information is the biggest;Whereas if problem information is identical with semanteme in answer information
Centre word the fewest, then answer information is the least with the value of the degree of association parameter of problem information.
It is pointed out that in other embodiments of the invention, the method is calculating the degree of association of answer information
The prediction used during parameter can also comprise other rule of reason, the invention is not restricted to this.
In the present embodiment, when after the degree of association parameter obtaining each answer information in answer information set, the method
According to the value size of degree of association parameter, each answer information in answer information aggregate can be ranked up, and select
Take forward N (value of N can be set according to actual needs) the individual answer information of sequence to be used as updating
Answer set, this renewal answer set is the alternative answer of problem information in target question and answer information, the most real
The now optimization to conversational system knowledge base.
It is pointed out that and utilize this method can be according to actual needs to update cycle of conversational system knowledge base
It is set (such as to be focused on by the question and answer information that every day is collected thus realizing every day pair
Conversational system knowledge base update is once), the invention is not restricted to this.Meanwhile, the method is upon execution, acquired
Mass data not only include user data, also include that question and answer evaluate system and the data of user evaluation.Tool
Body ground, these data include but not limited to: the problem information of user and the answer information of conversational system, dialogue system
The system source of answer information, the related data of user, the score data of answer information and the time of data genaration
Deng.
It can be seen from the above description that the conversational system knowledge base optimization method that the present embodiment is provided achieve right
The self study of telephone system knowledge base and closed loop feedback.Wherein, the method is so that conversational system can be to knowing
Know bad in storehouse and irrational question and answer information independence ground to carry out rationalizing and improve, so that the matter of this system
Amount can have the lifting of persistence.Meanwhile, the method can also make conversational system realize autonomic learning, so
The most just can depart from the manual learning style having to rely on attendant.
The method by the problem information in target question and answer information is pushed to again user obtain user for
The answer of this problem information, the most just achieves the closed loop feedback of " user-conversational system-user ", the most right
Telephone system the most just can keep the stability of self when by external interference.It is based on this closed-loop structure,
What conversational system can continue receives the feedback from user's perception aspect, and comes not according to the feedback of user
Disconnected optimize its knowledge storehouse thus adjust the output of self, so that this output can meet the expectation of user.
Such as the problem information " the most alone " in target question and answer information as above, existing
The answer information that conversational system is fed back is to make user's satisfied " I don't have the slightest idea about the meaning of your remarks ".
And after conversational system knowledge base is updated by the method utilizing the present embodiment to be provided, conversational system then can be defeated
Going out the answer information of such as " I accompanies with you always ", this answer information obviously more conforms to user's expectation.
Embodiment two:
Fig. 2 shows the flowchart of the optimization method of the conversational system knowledge base that the present embodiment provided.
As in figure 2 it is shown, the method that the present embodiment is provided obtains question and answer to be analyzed letter the most in step s 201
The test and appraisal parameter of breath, and in step S202, judge whether test and appraisal parameter meets default optimisation criteria, if symbol
Close, then in step S203 using question and answer information to be analyzed acquired in step S201 as target question and answer information,
Target question and answer information is the question and answer information needing to be optimized.
It is pointed out that in the present embodiment, step S201 to step S203 realize principle and realization
Process and step S101 in embodiment one to step S103 realize principle and to realize process identical, therefore at this
Repeat no more.
Existing conversational system does not differentiates between user when carrying out problem propelling movement, and this is it is possible to cause conversational system
Problem information can be pushed to unaccommodated user, the answer information that this kind of user is fed back is likely to cause and answers
Case unreliable information and inaccurate.
Such as, when conversational system has pushed problem information " cosmetics of X board are how " to a male user,
Male user generally cannot accurately provide the answer of this problem information, the most also it is possible to feed back all to conversational system
Such as " I does not knows " or the answer information of " I does not uses cosmetics ".
For the problems referred to above, the method that the present embodiment is provided preferably is determined by the user property of user asks
The targeted customer of topic information pushing, thus obtain answer information the most accurate, reliable.
In the present embodiment, when, after the target question and answer information determined in conversational system data base, the method can be mesh
Problem information in mark question and answer information is pushed to the specific user of conversational system, and records user to this problem information
Feedback, thus obtain the answer information aggregate corresponding to problem information.
Specifically, as in figure 2 it is shown, after determining target question and answer information, the method obtains in step S204
Take the attribute of different user in conversational system, and in step S205, judge use accessed in step S204
Whether family attribute meets default problem pushes requirement.If the user property of user meets default problem, propelling movement is wanted
Ask, then corresponding user is defined as targeted customer in step S206 by the party's rule, and to targeted customer
Push the problem information in target question and answer information.
In the present embodiment, for determining that whether user is that the user property of targeted customer preferably includes: Yong Hushen
Part information, customer position information, age of user information and user gender information etc..Such as age of user
For information, if the age of the age of user in target question and answer information involved by problem information and certain user is in
Same age bracket, then the attribute of this user the most just meets default problem and pushes requirement.
Certainly, in different embodiments of the invention, when the method judges in step S205, both may be used
To simply use a certain item in item listed above or a few item to carry out the differentiation of targeted customer, it is also possible to utilize
Other reasonable items not listed above or other reasonable items carry out mesh with the combination of above-mentioned a certain item or a few
The differentiation of mark user, the invention is not restricted to this.
After pushing the problem information in target question and answer information to targeted customer, the method is logical in step S207
Cross conversational system and obtain the answer information that targeted customer is fed back for problem information, thus obtain answer information collection
Close.
After obtaining answer information set, the method calculates in answer information set each in step S208 respectively
The degree of association parameter of individual answer information, and in step S209 degree of association parameter according to each answer information from
Answer information determines renewal answer set in combining.
It is pointed out that in the present embodiment, the realizing principle and realized of step S207 to step S209
Journey and step S104 in embodiment one to step S106 realize principle and to realize process similar, therefore at this not
Repeat again.
This problem in the present embodiment, when the answer information in target question and answer information is updated, after renewal
The answer information of information preferably will comprise multiple answer information, when such user and conversational system interact,
Conversational system to the multiple different answer information of user feedback, thus can avoid use for same problem information
Family conversational system when inquiring same problem makes user be fed up with owing to always exporting identical answer, this
Further increase Consumer's Experience and user's viscosity of conversational system.
It can be seen from the above description that the conversational system knowledge base optimization method that the present embodiment is provided is in embodiment
On the basis of one method provided, push target question and answer by the attribute according to user to specific user and believe
Problem information in breath.Attribute and problem letter in conversational system input target question and answer information due to this kind of user
The attribute of the user of breath is same or like, is therefore also obtained with being directed to target question and answer by this kind of user and believes
The higher answer information of the most accurate or compatible degree of problem information in breath, and determined according to this kind of answer information
The renewal answer set of the problems referred to above information gone out also will be the most accurate.
It is pointed out that in other embodiments of the invention, the method reasonably can also be joined according to other
Number (such as interaction scenarios and/or mutual topic) chooses targeted customer from multiple users of conversational system, this
Invention is not limited to this.The most in one embodiment of the invention, the holding of the optimization method of conversational system knowledge base
Row step S301 to S309 is roughly the same to step S209 with step S201 in embodiment two, except for the difference that,
In this embodiment, the friendship that the method is acquired in step s 304 the different user for conversational system is current
Topic mutually, and utilize mutual topic to choose targeted customer in step S305.
In concrete application process, if a certain user and be " football " to the mutual topic between change system,
And problem information is for " AlphaGo " in target question and answer information, then now problem information is pushed to
This user is clearly the loftiest, in this case, will be unable to ensure that this user is for the problem pushed
The reliability of the answer information that information is fed back.The propelling movement strategy of the most this problem information can make pushed
User feels that interaction is interfered, thus reduces the Consumer's Experience of conversational system.
And the method that the present embodiment is provided is exactly based on mutual topic and selects from a large number of users of conversational system
Take targeted customer, when current mutual topic and the topic phase involved by problem information in target question and answer information of user
Guan Shi, is now pushed to problem information this user and user not only will not be made to feel lofty, additionally it is possible to allow user feel
Feel that conversational system has " thought ".In this case, the answer letter that user is fed back for problem information
Ceasing obvious quality higher, it more can be sent out exactly and mirror user and truly answer this problem information.
It is pointed out that in other embodiments of the invention, the optimization method of this conversational system knowledge base can
So that multiple different parameters (attribute of such as user, mutual topic and interaction scenarios) are combined more smart
True chooses targeted customer, and the present invention is similarly not so limited to.
It is also desirable to it is noted that in the foregoing description, in the test and appraisal parameter obtaining question and answer information to be analyzed
Before, it is also possible to the question and answer information to be analyzed got is filtered, determines the legitimacy of question and answer information with this.
Present invention also offers the optimization device of a kind of conversational system knowledge base, Fig. 4 shows should in the present embodiment
The structural representation of device.
As shown in Figure 4, the conversational system knowledge base that the present embodiment is provided optimizes device and preferably includes: target
Question and answer data obtaining module 401 and renewal answer set generation module 402.Wherein, target question and answer acquisition of information
Module 401 user judges whether the test and appraisal parameter of question and answer information to be analyzed meets default optimisation criteria.If met,
Target question and answer data obtaining module 401 then using this question and answer information as target question and answer information, wherein target question and answer letter
Breath is question and answer information to be optimized.
In the present embodiment, the test and appraisal parameter accessed by target question and answer data obtaining module 401 is preferably question and answer
The scoring to question and answer information to be analyzed of the evaluation system.For the problem information in question and answer information, its answer is believed
The scoring of breath is the highest, and also just explanation user is the highest to the satisfaction of this answer information;Whereas if answer letter
The scoring of breath is the lowest, and also just explanation user is the lowest to the satisfaction of this answer information, and this kind of answer information is also
Need the information being optimized just.
Target question and answer data obtaining module 401 is if it is judged that the value of test and appraisal parameter of question and answer information to be analyzed
Less than presetting test and appraisal threshold value, then judgement test and appraisal parameter is met default excellent by target question and answer data obtaining module 401
Change standard, therefore target question and answer data obtaining module 401 also will judge that this question and answer information to be analyzed is as target question and answer
Information.
After determining target question and answer information, target question and answer information can be passed by target question and answer data obtaining module 401
It is defeated by renewal answer set generation module 402, with by updating answer set generation module 402 according to target question and answer
Information generates the renewal answer set of answer corresponding to problem information in more fresh target question and answer information.
Specifically, in the present embodiment, updating answer set generation module 402 can be target under specific scene
Problem information in question and answer information is pushed to the user of conversational system, and records anti-to this problem information of user
Feedback, thus obtain the answer information aggregate corresponding to problem information.
Update answer set generation module 402 and preferably determine that problem information pushes away by the user property of user
The targeted customer sent, thus obtain answer information the most accurate, reliable.Therefore, in the present embodiment, update
Answer set generation module 402, before carrying out problem propelling movement, can obtain the attribute of different user in conversational system,
And whether the user property accessed by judging meets default problem and pushes requirement.If the user property symbol of user
Close the problem of presetting and push requirement, then update answer set generation module 402 and then corresponding user is defined as mesh
Mark user, and push the problem information in target question and answer information to targeted customer.
In the present embodiment, for determining that whether user is that the user property of targeted customer preferably includes: Yong Hushen
Part information, customer position information, age of user information and user gender information etc..Certainly, the present invention's
In different embodiments, renewal answer set generation module 402, both can be only when choosing targeted customer
Only use a certain item in item listed above or a few item to carry out the differentiation of targeted customer, it is also possible to more than utilization
Other unlisted reasonable items or other reasonable items carry out target use with the combination of above-mentioned a certain item or a few
The differentiation at family, the invention is not restricted to this.
It is pointed out that in other embodiments of the invention, update answer set generation module 402 and also may be used
To carry out the multiple users from conversational system according to other rational parameters (such as interaction scenarios and/or mutual topic)
In choose targeted customer, the invention is not restricted to this.
The most in one embodiment of the invention, update being used for acquired in answer set generation module 402 to select
Take the mutual topic that the different user that parameter is conversational system of targeted customer is current.The side that the present embodiment is provided
Method is exactly based on mutual topic and chooses targeted customer from a large number of users of conversational system, when the current friendship of user
When topic is relevant to the topic involved by problem information in target question and answer information mutually, now problem information is pushed to
This user not only will not make user feel lofty, additionally it is possible to conversational system has " thought " to allow user feel.
In this case, the obvious quality of answer information that user is fed back for problem information is higher, and it more can be accurate
Really send out and mirror user this problem information is truly answered.
It is pointed out that in other embodiments of the invention, the optimization method of this conversational system knowledge base can
So that multiple different parameters (attribute of such as user, mutual topic and interaction scenarios) are combined more smart
True chooses targeted customer, and the present invention is similarly not so limited to.
It is also desirable to it is noted that in the foregoing description, in the test and appraisal parameter obtaining question and answer information to be analyzed
Before, it is also possible to the question and answer information to be analyzed got is filtered, determines the legitimacy of question and answer information with this.
After obtaining answer information set, it is mostly by multiple owing to answering the answer information included in information aggregate
Different users is fed back, and the most both may comprise the answer information that user can be made more to be satisfied with, also
May comprise and cannot make customer satisfaction system answer information, the most also be accomplished by answer information aggregate is screened,
To be determined to the information of the alternative answer of problem information in target question and answer information.
Therefore, after obtaining answer information set, update answer set generation module 402 and calculate answer letter respectively
The degree of association parameter of each answer information in breath set, and generate for target question and answer information according to degree of association parameter
The renewal answer information set of middle problem information.
It is pointed out that in the present embodiment, update answer set generation module 402 and calculate each answer information
The principle of degree of association parameter identical with the principle illustrated in the optimization method of above-mentioned conversational system knowledge base, therefore
Do not repeat them here.
It should be understood that disclosed embodiment of this invention is not limited to ad hoc structure disclosed herein, process
Step or material, and the equivalent that should extend to these features that those of ordinary skill in the related art are understood is replaced
Generation.It is to be further understood that term as used herein is only used for describing the purpose of specific embodiment, and and unexpectedly
Taste restriction.
Special characteristic that " embodiment " mentioned in description or " embodiment " mean to describe in conjunction with the embodiments,
Structure or characteristic are included at least one embodiment of the present invention.Therefore, description various places throughout occurs
Phrase " embodiment " or " embodiment " same embodiment might not be referred both to.
Although above-mentioned example is for illustrating present invention principle in one or more application, but for this area
For technical staff, in the case of without departing substantially from the principle of the present invention and thought, hence it is evident that can in form, use
In the details of method and enforcement, various modifications may be made and need not pay creative work.Therefore, the present invention is by appended power
Profit claim limits.
Claims (10)
1. the optimization method of a conversational system knowledge base, it is characterised in that including:
Target question and answer information acquiring step, it is judged that whether the test and appraisal parameter of question and answer information to be analyzed meets optimization mark
Standard, as met, then using described question and answer information to be analyzed as target question and answer information;
Updating answer set generation step, problem information based on described target question and answer information obtains corresponding answer
Information aggregate, calculates the degree of association parameter of each answer information in described answer information set respectively, and according to institute
State degree of association parameter to generate for the renewal answer set of problem information in described target question and answer information.
2. the method for claim 1, it is characterised in that at described target question and answer information acquiring step
In, it is judged that whether the test and appraisal parameter of described question and answer information to be analyzed is less than is preset test and appraisal threshold value, if it is less,
Using described question and answer information to be analyzed as target question and answer information.
3. method as claimed in claim 1 or 2, it is characterised in that generate in described renewal answer set
In step, determine according to the centre word whether problem information in question and answer information has semanteme identical with answer information
The degree of association parameter of answer information, wherein, the quantity of semantic identical centre word is the most, answer information and problem
The degree of association parameter of information is the biggest.
4. method as claimed in claim 1 or 2, it is characterised in that obtaining described answer information set
Time, choose targeted customer according to preset rules, and push in described target question and answer information to described targeted customer
Problem information, obtains the answer information that described targeted customer is fed back for described problem information, thus obtains institute
State question and answer information aggregate.
5. method as claimed in claim 4, it is characterised in that when choosing described targeted customer,
Obtain the user property of conversational system different user, it is judged that whether user property meets the propelling movement of default problem is wanted
Asking, if met, then corresponding user being defined as described targeted customer, wherein, described user property includes
Any one in item set forth below or several:
Subscriber identity information, customer position information, age of user information, user gender information;
Or, obtain the interaction scenarios of conversational system different user and/or mutual topic, it is judged that described interaction scenarios
And/or whether topic meets the propelling movement requirement of default problem alternately, if met, then corresponding user is defined as institute
State targeted customer.
6. the optimization device of a conversational system knowledge base, it is characterised in that including:
Target question and answer data obtaining module, it is for judging whether the test and appraisal parameter of question and answer information to be analyzed meets excellent
Change standard, as met, then using described question and answer information to be analyzed as target question and answer information;
Updating answer set generation module, it obtains correspondence for problem information based on described target question and answer information
Answer information set, calculate the degree of association parameter of each answer information in described answer information set respectively, and
Generate for the renewal answer set of problem information in described target question and answer information according to described degree of association parameter.
7. device as claimed in claim 6, it is characterised in that described target question and answer data obtaining module is joined
It is set to judge whether the test and appraisal parameter of described question and answer information to be analyzed is less than and presets test and appraisal threshold value, if it is less,
Using described question and answer information to be analyzed as target question and answer information.
Device the most as claimed in claims 6 or 7, it is characterised in that described renewal answer set generates mould
Block is configured to whether have the identical centre word of semanteme according to problem information in question and answer information with answer information to be come really
Determining the degree of association parameter of answer information, wherein, the quantity of semantic identical centre word is the most, answer information with ask
The degree of association parameter of topic information is the biggest.
Device the most as claimed in claims 6 or 7, it is characterised in that described renewal answer set generates mould
Block is configured to, when obtaining described answer information set, choose targeted customer according to preset rules, and to described mesh
Mark user pushes the problem information in described target question and answer information, obtains described targeted customer and believes for described problem
The answer information that breath is fed back, thus obtain described question and answer information aggregate.
10. device as claimed in claim 9, it is characterised in that described renewal answer set generation module is joined
It is set to when choosing described targeted customer,
Obtain the user property of conversational system different user, it is judged that whether user property meets the propelling movement of default problem is wanted
Asking, if met, then corresponding user being defined as described targeted customer, wherein, described user property includes
Any one in item set forth below or several:
Subscriber identity information, customer position information, age of user information, user gender information;
Or, obtain the interaction scenarios of conversational system different user and/or mutual topic, it is judged that described interaction scenarios
And/or whether topic meets the propelling movement requirement of default problem alternately, if met, then corresponding user is defined as institute
State targeted customer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610262475.7A CN105912697B (en) | 2016-04-25 | 2016-04-25 | A kind of optimization method and device of conversational system knowledge base |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610262475.7A CN105912697B (en) | 2016-04-25 | 2016-04-25 | A kind of optimization method and device of conversational system knowledge base |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105912697A true CN105912697A (en) | 2016-08-31 |
CN105912697B CN105912697B (en) | 2019-08-27 |
Family
ID=56751992
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610262475.7A Active CN105912697B (en) | 2016-04-25 | 2016-04-25 | A kind of optimization method and device of conversational system knowledge base |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105912697B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106446141A (en) * | 2016-09-21 | 2017-02-22 | 北京光年无限科技有限公司 | Interaction data processing method for intelligent robot system and robot system |
CN106682137A (en) * | 2016-12-19 | 2017-05-17 | 武汉市灯塔互动文化传播有限公司 | Intelligent stock investment adviser questioning-answering interaction method and intelligent stock investment adviser questioning-answering interaction system |
CN106781790A (en) * | 2016-12-10 | 2017-05-31 | 杭州博世数据网络有限公司 | A kind of on-line study system with self-picketing orthofunction |
CN107729510A (en) * | 2017-10-23 | 2018-02-23 | 深圳市前海众兴电子商务有限公司 | Information interacting method, information interactive terminal and storage medium |
CN109460459A (en) * | 2018-10-31 | 2019-03-12 | 神思电子技术股份有限公司 | A kind of conversational system automatic optimization method based on log study |
CN109657038A (en) * | 2018-10-10 | 2019-04-19 | 阿里巴巴集团控股有限公司 | The method for digging, device and electronic equipment of a kind of question and answer to data |
CN110046235A (en) * | 2019-03-18 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of knowledge base appraisal procedure, device and equipment |
CN110263051A (en) * | 2019-06-11 | 2019-09-20 | 出门问问信息科技有限公司 | Question and answer for question answering system are to update method, device, equipment and storage medium |
CN110674276A (en) * | 2019-09-23 | 2020-01-10 | 深圳前海微众银行股份有限公司 | Robot self-learning method, robot terminal, device and readable storage medium |
CN111382239A (en) * | 2018-12-27 | 2020-07-07 | 上海智臻智能网络科技股份有限公司 | Method and device for optimizing interaction flow |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102637170A (en) * | 2011-02-10 | 2012-08-15 | 北京百度网讯科技有限公司 | Question pushing method and system |
CN103365899A (en) * | 2012-04-01 | 2013-10-23 | 腾讯科技(深圳)有限公司 | Question recommending method and question recommending system both in questions-and-answers community |
CN103425640A (en) * | 2012-05-14 | 2013-12-04 | 华为技术有限公司 | Multimedia questioning-answering system and method |
CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
CN104050256A (en) * | 2014-06-13 | 2014-09-17 | 西安蒜泥电子科技有限责任公司 | Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method |
US20160092523A1 (en) * | 2014-09-30 | 2016-03-31 | International Business Machines Corporation | Information handling system and computer program product for dynamcally assigning question priority based on question extraction and domain dictionary |
CN105512228A (en) * | 2015-11-30 | 2016-04-20 | 北京光年无限科技有限公司 | Bidirectional question-answer data processing method and system based on intelligent robot |
-
2016
- 2016-04-25 CN CN201610262475.7A patent/CN105912697B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102637170A (en) * | 2011-02-10 | 2012-08-15 | 北京百度网讯科技有限公司 | Question pushing method and system |
CN103365899A (en) * | 2012-04-01 | 2013-10-23 | 腾讯科技(深圳)有限公司 | Question recommending method and question recommending system both in questions-and-answers community |
CN103425640A (en) * | 2012-05-14 | 2013-12-04 | 华为技术有限公司 | Multimedia questioning-answering system and method |
CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
CN104050256A (en) * | 2014-06-13 | 2014-09-17 | 西安蒜泥电子科技有限责任公司 | Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method |
US20160092523A1 (en) * | 2014-09-30 | 2016-03-31 | International Business Machines Corporation | Information handling system and computer program product for dynamcally assigning question priority based on question extraction and domain dictionary |
CN105512228A (en) * | 2015-11-30 | 2016-04-20 | 北京光年无限科技有限公司 | Bidirectional question-answer data processing method and system based on intelligent robot |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106446141B (en) * | 2016-09-21 | 2020-05-19 | 北京光年无限科技有限公司 | Interactive data processing method for intelligent robot system and robot system |
CN106446141A (en) * | 2016-09-21 | 2017-02-22 | 北京光年无限科技有限公司 | Interaction data processing method for intelligent robot system and robot system |
CN106781790A (en) * | 2016-12-10 | 2017-05-31 | 杭州博世数据网络有限公司 | A kind of on-line study system with self-picketing orthofunction |
CN106682137A (en) * | 2016-12-19 | 2017-05-17 | 武汉市灯塔互动文化传播有限公司 | Intelligent stock investment adviser questioning-answering interaction method and intelligent stock investment adviser questioning-answering interaction system |
CN107729510A (en) * | 2017-10-23 | 2018-02-23 | 深圳市前海众兴电子商务有限公司 | Information interacting method, information interactive terminal and storage medium |
CN107729510B (en) * | 2017-10-23 | 2021-07-06 | 深圳市前海众兴科研有限公司 | Information interaction method, information interaction terminal and storage medium |
CN109657038A (en) * | 2018-10-10 | 2019-04-19 | 阿里巴巴集团控股有限公司 | The method for digging, device and electronic equipment of a kind of question and answer to data |
CN109460459A (en) * | 2018-10-31 | 2019-03-12 | 神思电子技术股份有限公司 | A kind of conversational system automatic optimization method based on log study |
CN111382239A (en) * | 2018-12-27 | 2020-07-07 | 上海智臻智能网络科技股份有限公司 | Method and device for optimizing interaction flow |
CN111382239B (en) * | 2018-12-27 | 2023-06-23 | 上海智臻智能网络科技股份有限公司 | Interaction flow optimization method and device |
CN110046235A (en) * | 2019-03-18 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of knowledge base appraisal procedure, device and equipment |
CN110263051A (en) * | 2019-06-11 | 2019-09-20 | 出门问问信息科技有限公司 | Question and answer for question answering system are to update method, device, equipment and storage medium |
CN110674276A (en) * | 2019-09-23 | 2020-01-10 | 深圳前海微众银行股份有限公司 | Robot self-learning method, robot terminal, device and readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN105912697B (en) | 2019-08-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105912697A (en) | Optimization method and device of dialog system knowledge base | |
CN106791235B (en) | A kind of method, apparatus and system selecting service agent | |
CN106845624A (en) | The multi-modal exchange method relevant with the application program of intelligent robot and system | |
CN103916306B (en) | The method and apparatus for realizing instant messaging | |
CN108351986A (en) | Learning system, learning device, learning method, learning program, training data generating means, training data generation method, training data generate program, terminal installation and threshold value change device | |
CN104865927A (en) | Method and apparatus for constructing user model, and equipment control method and apparatus | |
CN109446387A (en) | A kind of Owners Committee's intelligent Answer System based on artificial intelligence | |
CN110413748A (en) | Based on big data robot service knowledge base entry recommender system | |
CN106375369A (en) | Mobile Web service recommendation method and collaborative recommendation system based on user behavior analysis | |
CN107122447A (en) | The network searching system and control method of a kind of multi-data source fusion based on preference | |
CN105808590A (en) | Search engine realization method as well as search method and apparatus | |
CN104007805A (en) | Method and device for achieving power saving of mobile terminal and mobile terminal | |
CN111581516A (en) | Investment product recommendation method and related device | |
CN104732450A (en) | Power supply quality evaluation system based on rule engine | |
CN107623621A (en) | Language material collection method of chatting and device | |
KR102117287B1 (en) | Method and apparatus of dialog scenario database constructing for dialog system | |
JP2022003586A (en) | Supporting system | |
CN105472413A (en) | Television channel sorting method and system, and server | |
CN101136869A (en) | Method for generating search intention based contacts group of instant communication system | |
CN106503189A (en) | search system optimization method and device based on artificial intelligence | |
CN113163063B (en) | Intelligent outbound call system and method | |
CN113420058B (en) | Conversational academic conference recommendation method based on combination of user historical behaviors | |
CN102880725B (en) | Recommending method based on demand-based service organization | |
Huang et al. | Applied to mobile multimedia intelligent speech system interactive topic guiding model | |
CN106874451A (en) | A kind of method of the personal exclusive corpus of automatic foundation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |