CN103729424B - Evaluation method and system is answered in Ask-Answer Community - Google Patents
Evaluation method and system is answered in Ask-Answer Community Download PDFInfo
- Publication number
- CN103729424B CN103729424B CN201310714726.7A CN201310714726A CN103729424B CN 103729424 B CN103729424 B CN 103729424B CN 201310714726 A CN201310714726 A CN 201310714726A CN 103729424 B CN103729424 B CN 103729424B
- Authority
- CN
- China
- Prior art keywords
- answer
- weight
- feature
- information
- 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.)
- Active
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The invention provides answering evaluation method in a kind of Ask-Answer Community, the method includes:Corresponding all answer contents and the multidimensional information related to the answer content under acquisition problem and the problem;Based on the basic weight for calculating each answer content to the mode that each dimension information carries out individual weighting;Influencing each other for comprehensive each dimension information, determines the corresponding basic weight for adjusting power mechanism to adjust the acquisition, obtains the final weight of each answer content;Overall merit is carried out to all answer contents based on the final weight.A kind of Ask-Answer Community in answer evaluation system is correspondingly additionally provided.The method and system that the present invention is provided can effectively screen the Consumer's Experience for lifting answer platform to the valuable answer of problem.
Description
Technical field
The present invention relates to answering evaluation method and system in computer network field, more particularly to a kind of Ask-Answer Community.
Background technology
At present, the important channel that related information is that user obtains information is searched for by search platform, especially in question and answer society
Search problem, proposition problem in area, answer a question, browse problem or additional problem etc., this has become carry out between user interactive letter
The important way of breath exchange.Wherein, common Ask-Answer Community have Baidu to know, search ask, Sina love ask.
Generally, in Ask-Answer Community, under each problem, the displaying order of answer content is based primarily upon following two modes:1)Only press
It is ranked up according to the time that answers a question, i.e., the answer of displaying in the top is in time closer to the time of current search;2)
Favorable comment number according to acquisition is answered is ranked up, i.e., the approval number that the answer under same problem obtains user is more, and which is more forward
It is illustrated in Ask-Answer Community.But, both modes respectively have which not enough, for first kind of way, due to returning for forward displaying
Answer and be not necessarily the answer for most mating the problem, therefore, user generally needs to take a long time the answer required for finding, and
And, with being incremented by for number is answered, its deficiency is more obvious for this mode;For the second way, several to returning based on approving of
Answer and be ranked up, this is easy to suffer spam(Electronic waste)The attack of user so that those are directly beneficial to spam user
Ad content top to forward display location, so as to cause the misleading of the user to browsing the answer.
Content of the invention
It is an object of the invention to provide evaluation method and system is answered in a kind of Ask-Answer Community, can effectively lift question and answer and put down
The Consumer's Experience of platform.
According to an aspect of the invention, there is provided answering evaluation method in a kind of Ask-Answer Community, the method includes:
Corresponding all answer contents and the multidimensional information related to the answer content under acquisition problem and the problem;
By including regression model, based on the base for calculating each answer content to the mode that each dimension information carries out individual weighting
This weight;
Influencing each other for comprehensive each dimension information, determines the corresponding basic weight for adjusting power mechanism to adjust the acquisition, obtains
The final weight of each answer content;
Overall merit is carried out to all answer contents based on the final weight.
According to another aspect of the present invention, answer evaluation system in a kind of Ask-Answer Community is additionally provided, including:
Information acquisition unit, for obtain under problem and the problem corresponding all answer contents and with the answer in
Hold related multidimensional information;
Basic weight calculation unit, by including regression model, based on the mode meter for carrying out individual weighting to each dimension information
Calculate the basic weight of each answer content;
Weight adjustment unit, for influencing each other for comprehensive each dimension information, determines
The basic weight for taking, obtains the final weight of each answer content;
Evaluation unit is answered, overall merit is carried out to all answer contents based on the final weight.
Compared with prior art, the present invention has advantages below:
1) present invention is effectively screened to the valuable answer of problem by the assessment to answering information, and will be excellent for the answer
Viewer and quizmaster is first presented to, the Consumer's Experience of answer platform is improved;
2) present invention can be effectively prevented junk information(spam)The attack of user, it is to avoid the category information is to browsing user
Cause to mislead.
Description of the drawings
By reading the detailed description made by non-limiting example that is made with reference to the following drawings, other of the invention
Feature, objects and advantages will become more apparent upon:
Fig. 1 is to answer evaluation method flow chart in Ask-Answer Community according to a preferred embodiment of the invention;
Fig. 2 is the length of the answer content shown in the present embodiment and the corresponding curve map for adjusting weight coefficient;
Fig. 3 is the graph of relation according to the user gradation of the preferred embodiment of the present invention and the quality of answer content;
The schematic block diagrams of answering community in answer evaluation system of the Fig. 4 for another preferred embodiment of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings the present invention is described in further detail.
According to an aspect of the invention, there is provided answering evaluation method in a kind of Ask-Answer Community.It should be noted that under
The weight that text is mentioned with answer information quality into positive relationship, weight is higher, and the quality for representing answer information is more excellent.Described time
The quality of information is answered mainly from answer content, the user behavior feature of the submission answer, the user characteristics for browsing the answer
Comprehensive measurement is carried out etc. information.
Refer to Fig. 1, Fig. 1 is evaluation method flow process to be answered in Ask-Answer Community according to a preferred embodiment of the invention
Figure.
As shown in figure 1, method provided by the present invention is comprised the following steps:
Step S101, obtains under problem and the problem corresponding all answer contents and related to the answer content
Multidimensional information.
Specifically, in order to preferably evaluate to the value of the answer information in Ask-Answer Community, asked based on Network Capture
Corresponding all answer contents and relevant information under all problems and the problem is answered in community, the concrete mode here for obtaining is not
It is restricted.
The multidimensional information related to the answer content mainly includes:The characteristic information for answering itself, submission are described
The user's characteristic information of answer content, the user behavior characteristic information for browsing the problem and answer.Wherein, the answer itself
Characteristic information include non-textual characteristic information and text feature information;The user's characteristic information for submitting the answer content to
Rate is adopted including user gradation and user;The user behavior feature for browsing the problem and answer refers mainly to the user to returning
The evaluation information that answers, for example common answer favorable comment number, answer in thank you and language and question closely language of thanking you in answering etc., should
Information can portray feedback information of the user to the answer.
Wherein, the text feature information that answers in the characteristic information of itself mainly includes:Special marking feature, core
Presentive word feature, query tendency feature and meaningless feature, tendentiousness of thanking you feature.
Wherein, the non-textual characteristic information that answers in the characteristic information of itself mainly includes:The answer content
Length information, the paragraph number of the answer content, Rich Media's characteristic information or/and question closely and answer information.Wherein, Rich Media is special
Reference breath refers mainly to the characteristic informations such as picture in answer content, map.
Step S102, by including regression model, calculates each time based on to the mode that each dimension information carries out individual weighting
Answer the basic weight of content.
Specifically, after obtaining above-mentioned multidimensional information, based on the quality that above-mentioned multidimensional information weighs each answer content.More
Body ground, the basic weight for calculating each answer content by following computing formula includes regression model, to each dimension information
Carry out linear weighted function calculating.Computing formula is as follows:
scoreini=radio1×dimesion1+…radioi×dimesioni+…radion×dimesionn
Wherein, radio1、radioi、radionThe tune weight factor of Ge Wei information, dimesion are represented respectively1、
dimesioni、dimesionnThe weight of Ge Wei information, score are represented respectivelyiniRepresent basic weight.Wherein, described adjust power because
Son and weight can be by sides such as the exhaustion to feature and main characteristic information included in each dimension information, selection, cure parameters
Method is determined.
Each dimension information is referred mainly to:The characteristic information for answering itself, the user characteristics for submitting the answer content to
Information, the user behavior characteristic information for browsing the problem and answer.
By the calculating of above-mentioned formula, the basic weight of each answer content can be obtained, at tune power hereafter
Reason, you can to obtain the final weight of each answer content.
Step S103, influencing each other for comprehensive each dimension information, determines
Weight, obtains the final weight of each answer content.
Specifically, influencing each other for each dimension information refers mainly to each dimension information to front produced by the quality of answer content
Or negative impact.The tune power mechanism is then weighted or drops power according to the just negative, degree of influence of the impact, i.e., every
Corresponding tune weight coefficient is determined in the basic weight of individual answer content, and including weighting or dropping weight coefficient, both products are most
The weight for obtaining eventually.Specific computing formula is referred to hereafter:
score=w1×…Wm×scoreini
Wherein, w1、wmRepresent and adjust weight coefficient, scoreiniRepresent that basic weight, score represent final weight.
Wherein, the weighting or drop weight coefficient are related to the specific features in each dimension information, hereafter will be discussed in more detail.
For the non-textual characteristic information in the above-mentioned answer characteristic information of itself, weighting or drop weight coefficient is wherein affected
Feature mainly include answer content length and answer content paragraph.
For respectively, due to the length of answer content in Ask-Answer Community be typically all moderate-length content its quality compared with
Height, the information that the content of short length excessively is generally comprised are more unilateral, and its quality is relatively low, and long content is then usually because tediously long and scarce
Weary keynote message, therefore, the length of answer content presents to the contribution of answer content quality and first increases the trend for reducing afterwards.For
The relation between length and answer content quality is preferably embodied, the relation curve multiple gears can be divided into and be carried out table
Show, it is possible to calculate the tune weight coefficient corresponding to the length of the answer content using equation below:
Wherein, len represents the hypothesis length of answer content, and 1 to n represents tune weight coefficient w respectively1Arrive wnCorresponding gear,
len1Arrive lennRepresent that 1 arrives the corresponding length of interval of n shelves, w respectivelyLengthRepresent that the length institute of the answer content for finally giving is right
The tune weight coefficient that answers.
Further, Fig. 2 is referred to, and Fig. 2 is that the length of the answer content shown in the present embodiment adjusts weight coefficient with corresponding
Curve map.As shown in Fig. 2 weighted calculation side of the calculating of the tune weight coefficient of the answer content length using above-mentioned multiple gears
Formula, the weight coefficient of adjusting between adjacent gear positions are obtained using adjust weight coefficient and the phase Calais of variable of adjacent low-grade location, are finally given
Adjust weight coefficient and the corresponding relation of content-length adopt such as (len1,w1),(len2,w2),(len3,w3)……(lenn,wn) etc.
With(Length, tune weight coefficient)Form representing.
Further, as described above, the paragraph number of answer content is also closely related with tune weight coefficient, and it can embody and answer
The structurized fine or not degree of case text, specifically can be using the corresponding tune of the paragraph of linearly increasing form calculus varying number
Weight coefficient, refers to equation below:
Wherein, pradioIt is that paragraph adjusts power radix;pnumIt is current answer paragraph number;ptopIt is the answer paragraph number threshold for setting
Value, wParagraphFor the corresponding tune weight coefficient of the calculated paragraph.
Further, for the Rich Media's characteristic information in the non-textual characteristic information, i.e. answer content includes all
Such as the characteristic information such as picture, map, then be directly weighted in the basic weight of the answer content.Equally, for including
The answer content of information of answering is questioned closely, according to length information, paragraph number and the Rich Media's feature that questions closely and answer content
Information etc. is accordingly adjusted power to process.
For submitting the user's characteristic information of the answer content to, the feature master of wherein affected weighting or drop weight coefficient
User gradation to be included and user adopt rate.
In respectively, usual user gradation is higher, and the high possibility of the answer content quality is higher, but arrives certain journey
Degree, then gradually gentle, Fig. 3 is referred to, Fig. 3 is the quality of the user gradation according to the preferred embodiment of the present invention and answer content
Graph of relation, as shown in figure 3, the quality of answer content first increase sharply again as presents in the growth of user gradation gently progressive
Variation tendency, can adopt equation below(That is the form of Logarithmic calculation)Calculate the tune weight coefficient corresponding to user gradation:
Wherein, levelradioRepresent user gradation weighted factor, ulevelRepresent the grade of the user, toplevelIt is to set
Highest user gradation, wUser gradationRepresent the corresponding final tune weight coefficient of the user gradation.
Further, user adopts the probability that the answer of the i.e. user of rate is adopted, and which can be weighed the user's history and answer
The quality of content, according to the situation that the user's history answer content is adopted, can predict user's contribution high-quality answer
Possibility, can adopt equation below to this(That is the form of Logarithmic calculation)Calculate the user and adopt the corresponding tune weight coefficient of rate:
Wherein, good_rateradioRepresent user and adopt rate weighted factor;Good_rate represents the user and adopts rate;
topgood_rateIt is that the highest user for setting adopts rate value, wAdopt rateRepresent that the user adopts the corresponding final tune weight coefficient of rate.
For browsing the user behavior characteristic information of the problem and answer, the spy of weighting or drop weight coefficient is wherein affected
The tendentiousness feature of main language of thanking you including answer favorable comment number, user is levied, and questions closely the tendentiousness feature etc. of thanking you for answering the inside.
Wherein, a feature of the favorable comment number as user behavior characteristic information is answered, anti-to answered for portraying user
Feedforward information, the good evaluating data of certain answer of the major embodiment user to seeing, the relation object between this feature and tune weight coefficient
The corresponding relation that above-mentioned user is adopted between rate and tune weight coefficient is similar to, i.e., can be equally letter in the way of using Logarithmic calculation
For the sake of bright, here is not repeated.
For the text feature information that answers in the characteristic information of itself, wherein affect weighting or drop weight coefficient
Feature mainly includes that special marking feature, core presentive word feature, query tendency feature and meaningless feature, tendentiousness of thanking you are special
Levy.Hereafter put up with this four features to be described in detail.
For answer content is included such as《》、“”、<>Deng special marking feature, then carry out respective weight process.
Wherein, the core presentive word refer to reverse document-frequency weight exceed certain threshold value and through stop words, symbol,
The core word of the filtrations such as short number word alphabetic string.In the present embodiment, the analysis master in power mechanism is being adjusted to the core presentive word
To include two steps:1)Generate core vocabulary;2)Coupling core word.
Specifically, with regard to step 1)For, the main word frequency information passed through in statistical problem title is simultaneously filtered(For example cross
Filter stop words therein, symbol, short number word alphabetic string etc.), calculate the idf of word(Reverse document-frequency, inverse document
frequency)It is distributed and is formed the vocabulary comprising reverse document-frequency weight information.
With regard to step 2)For, being broadly divided into several steps as follows is carried out:
ⅰ)Certain threshold value is set, weight in the problem title is extracted and is more than the word of the threshold value and is arranged by weight
Sequence, retains most 2 core words in the top(Referred to as word 1, word 2);
ⅱ)In the core vocabulary of the formation, expand the synonym of institute's predicate 1 or/and word 2;
ⅲ)The adjustment institute predicate 1, weight of word 2, if the idf weights difference of two words is larger, carries out drop power to word 2
Process, to project the impact of the strong keyword of competency;
ⅳ)Fetch and answer most top n bytes and mated with the core word of the reservation, and the idf weights point by coupling
Smooth being mapped to of shelves is specified on interval, to avoid the noise of the rear partial content of long answer from producing shadow to the coupling of core word
Ring.
Wherein, the query tendency feature and meaningless feature are referred in answer content comprising tendency or the content sheet of having a question
The insignificant situation of body.Generally, in answer content, why the band tendency that has a question is because that problem is unclear.From the point of view of citing:
(1)Problem:Trade mark registration is invalid what if?
Answer:This must see that your trade mark is that what reason is invalid.
(2)Problem:How much this needs across web game to act on behalf of angle road?What is specifically needed?
Answer:Online friend needs any agency?
(3)Problem:May I ask why my excel becomes following two icon, have changed unfolding mode and also do not use?
Answer:Such icon was not met, shyly, be can't help busy.
From the point of view of by above three example,(1)With(2)In answer content belong to comprising the situation of tendency of having a question,(3)In
Answer content then belong to the insignificant situation of content itself.
For the answer for being inclined to feature and meaningless feature comprising the query, the form that is mainly mated by vocabulary,
In the range of limited replylen, power process is dropped in hit crucial word string accordingly.
Wherein, the tendentiousness feature of thanking you includes forward direction of the user in answer, negative sense or other kinds of evaluation letter
Breath.When analyzing the relation between this feature and tune weight coefficient, first, the words and phrases frequency of thanking you in being answered by statistics, and by all
Such as manual reviews(review)Mode obtain the tendentiousness dictionary for substantially characterizing positive and negative evaluation information;Secondly, mate tendentiousness
Dictionary, carries out tendency sex determination according to principle of the positive evaluation and optimization in negative sense evaluation, if it is decided that result positive for hit
Vocabulary, then be weighted;Hit negative sense vocabulary, then be not weighted;Otherwise, for the answer does not include the tendentiousness dictionary
The situation of information, then carry out respective weight process based on other situations mentioned above.
Generally speaking, based on the basis of mentioned above, weighting or drop power mechanism in the present embodiment also include following feelings
Shape:
The weight for answering the characteristic information of itself is too low, drop power;
Submit to the weight of the user's characteristic information of the answer content too low, drop power;
Answer is to recommend answer, best answers etc., weighting;
The short vocabulary or phrase that answers comprising special marking, weighting;
Situation is answered for questioning closely, different weightings are carried out according to different ratios.
Based on the mode that the above-mentioned basic weighting that enumerates is weighed with drop, on the basis of the initial answer weight for calculating, to returning
Answering carries out corresponding weighting and the process of drop power, generates the final weight of the answer content.
All answer contents are carried out overall merit based on the final weight by step S104.
Specifically, all answer contents are ranked up according to the final weight, answer content in the top
Best answers are then evaluated as, and the answer content for ranking behind then is evaluated as suboptimum answer, and is preferably existed according to sequence
Show the answer content and relevant information on the page from high to low.
Compared with prior art, method provided by the present invention has advantages below:Answer according to the value to puing question to certainly
The method of dynamic sequence is caused preferentially to represent answer valuable to problem and is possibly realized, and the method can optimize millions count issue
The sortord of lower answer so that browse user priority and see to the more helpful answer of solve problem, browse use so as to reduce
Family meets the time of demand after the page is reached and searches energy cost, optimizes viewing experience, is lifted and browses satisfaction.
According to another aspect of the present invention, answer evaluation system in a kind of Ask-Answer Community is additionally provided, Fig. 4 is refer to,
The schematic block diagrams of answering community in answer evaluation system of the Fig. 4 for another preferred embodiment of the present invention.As shown in figure 4, should
System includes:
Information acquisition unit 401, for obtain under problem and the problem corresponding all answer contents and with described time
Answer the related multidimensional information of content;
Basic weight calculation unit 402, by including regression model, based on the mode for carrying out individual weighting to each dimension information
Calculate the basic weight of each answer content;
Weight adjustment unit 403, for influencing each other for comprehensive each dimension information, it is described that determination adjusts power mechanism to adjust accordingly
The basic weight for obtaining, obtains the final weight of each answer content;
Evaluation unit 404 is answered, overall merit is carried out to all answer contents based on the final weight.
Below, the course of work of each unit provided by the present invention is specifically described.
Specifically, in order to preferably evaluate to the value of the answer information in Ask-Answer Community, described information obtains single
First 401 based on corresponding all answer contents and relevant information under all problems in Network Capture Ask-Answer Community and the problem,
This is not restricted for the concrete mode for obtaining.The multidimensional information related to the answer content mainly includes:The answer itself
Characteristic information, submit the user's characteristic information of the answer content to, browse the user behavior feature letter of the problem and answer
Breath.Wherein, the answer characteristic information of itself includes non-textual characteristic information and text feature information;Return described in the submission
The user's characteristic information for answering content includes that user gradation and user adopt rate;The user behavior for browsing the problem and answer
Feature refers mainly to the user to the evaluation information answered, such as thanking you in common answers favorable comment number, answer and chases after language
Language of thanking you for chasing after in answering etc. is asked, the information can portray feedback information of the user to the answer.
Wherein, the text feature information that answers in the characteristic information of itself mainly includes:Special marking feature, core
Presentive word feature, query tendency feature and meaningless feature, tendentiousness of thanking you feature.
Wherein, the non-textual characteristic information that answers in the characteristic information of itself mainly includes:The answer content
Length information, the paragraph number of the answer content, Rich Media's characteristic information or/and question closely and answer information.Wherein, Rich Media is special
Reference breath refers mainly to the characteristic informations such as picture in answer content, map.
After obtaining above-mentioned multidimensional information, based on the quality that above-mentioned multidimensional information weighs each answer content, and by weighing substantially
Re-computation unit 402 calculates the basic weight of each answer content by following computing formula, that is, include regression model, to described
Each dimension information carries out linear weighted function calculating.Computing formula is as follows:
scoreini=radio1×dimesion1+…radioi×dimesioni+…radion×dimesionn
Wherein, radio1、radioi、radionThe tune weight factor of Ge Wei information, dimesion are represented respectively1、
dimesioni、dimesionnThe weight of Ge Wei information, score are represented respectivelyiniRepresent basic weight.Wherein respectively dimension information is main
Refer to:The characteristic information for answering itself, the user's characteristic information for submitting the answer content to, browse the problem and answer
User behavior characteristic information.By the calculating of above-mentioned formula, the basic weight of each answer content can be obtained.
Wherein, the comprehensive each dimension information of the weight adjustment unit 403 is to front or negative produced by the quality of answer content
Impact.Corresponding weighting or drop power mechanism is determined to adjust the basic weight of the acquisition, specifically, in each answer content
Basic weight on determine that corresponding weighting or drop weight coefficient, both products are the final weight for obtaining.Specific calculating
Formula is referred to hereafter:
score=w1×…wm×scoreini
Wherein, w1、wmRepresent and adjust weight coefficient, scoreiniRepresent that basic weight, score represent final weight.Wherein, described
Weighting or drop weight coefficient are related to the specific features in each dimension information.As the relation between each feature and tune weight coefficient is as above
Described, for simplicity's sake, no longer describe in detail.
Wherein, the answer evaluation unit 404 all answer contents are ranked up according to the final weight and
Evaluate, answer content in the top is evaluated as best answers, the answer content for ranking behind is evaluated as suboptimum answer.
Preferably, the system also includes display unit, for the sequence according to final weight, is opened up on the page from high to low
Show the answer content and relevant information.
The system provided by the present invention has advantages below:The system is adjusted by basic weight calculation unit and weight
The process of unit, can preferably pick out the answer for having higher-value to problem, it is possible to be ordered as user according to be worth
Lift the experience of answer platform.
Above disclosed is only presently preferred embodiments of the present invention, can not limit certainly the right of the present invention with this
Scope, the equivalent variations that is therefore made according to the claims in the present invention still belong to the scope covered by the present invention.
Claims (14)
1. the answer evaluation method in a kind of Ask-Answer Community, the method include:
A) corresponding all answer contents and the multidimensional information related to the answer content under problem and the problem is obtained;
B) based on the basic weight for calculating each answer content to the mode that each dimension information carries out individual weighting;
C) influencing each other for information is comprehensively respectively tieed up, determines the basic weight of the corresponding tune power mechanism regulation acquisition, obtain every
The final weight of individual answer content;
D) overall merit is carried out to all answer contents based on the final weight;
Wherein, the multidimensional information mainly includes:The characteristic information for answering itself, the user spy for submitting the answer content to
Reference breath, the user behavior characteristic information for browsing the problem and answer;
Wherein, described answer the characteristic information of itself include special marking feature, core presentive word feature, query tendency feature and
Meaningless feature, tendentiousness of thanking you feature;
Wherein, the tune power mechanism is specifically included:
For the special marking feature, process is directly weighted;
For the core presentive word feature, then by generating, core vocabulary is corresponding with coupling core word determination to adjust weight coefficient;
Feature and meaningless feature are inclined to for the query, the form that mates by vocabulary, in limited replylen scope
Interior, the crucial word string of hit, the drop power for carrying out correlation are processed;
For the tendentiousness feature of thanking you, by obtaining the tendentiousness dictionary that characterizes evaluation information and by answer content and institute
State tendentiousness dictionary to be mated, carry out corresponding weighting process.
2. answer evaluation method according to claim 1, wherein, the answer characteristic information of itself includes answer content
Length and answer content paragraph number.
3. answer evaluation method according to claim 2, wherein, the tune power mechanism is specifically included:
For the length of answer content, corresponding tune weight coefficient is determined using the linear weighted function of multiple gears;
For the paragraph number of answer content, using the tune weight coefficient corresponding to linearly increasing form calculus.
4. answer evaluation method according to claim 1, wherein, the user's characteristic information for submitting the answer content to
Rate is adopted including user gradation and user.
5. answer evaluation method according to claim 4, wherein, the tune power mechanism is specifically included:
Rate is adopted for user gradation and user, the tune weight coefficient being respectively adopted corresponding to corresponding Logarithmic calculation form calculus.
6. the answer evaluation method according to any one of claim 1-5, wherein, the tune power mechanism also includes:
If the weight for answering the characteristic information of itself is too low, drop power;
If submitting to the weight of the user's characteristic information of the answer content too low, drop power;
If it is to recommend answer, best answers etc. to answer, weighting;
If vocabulary of the short answer content comprising special marking or phrase, weighting;
Situation is answered for questioning closely, different weightings are carried out according to different ratios.
7. the answer evaluation method according to any one of claim 1-5, wherein, the step b) is specifically included:
By including regression model, based on the basic power for calculating each answer content to the mode that each dimension information carries out individual weighting
Weight.
8. the answer evaluation system in a kind of Ask-Answer Community, including:
Information acquisition unit, for obtain under problem and the problem corresponding all answer contents and with the answer content phase
The multidimensional information of pass;
Basic weight calculation unit, based on the basic power for calculating each answer content to the mode that each dimension information carries out individual weighting
Weight;
Weight adjustment unit, for influencing each other for comprehensive each dimension information, determines
Basic weight, obtains the final weight of each answer content;
Evaluation unit is answered, overall merit is carried out to all answer contents based on the final weight;
Wherein, the multidimensional information mainly includes:The characteristic information for answering itself, the user spy for submitting the answer content to
Reference breath, the user behavior characteristic information for browsing the problem and answer;
Wherein, described answer the characteristic information of itself include special marking feature, core presentive word feature, query tendency feature and
Meaningless feature, tendentiousness of thanking you feature;
Wherein, the tune power mechanism is specifically included:
For the special marking feature, process is directly weighted;
For the core presentive word feature, then by generating, core vocabulary is corresponding with coupling core word determination to adjust weight coefficient;
Feature and meaningless feature are inclined to for the query, the form that mates by vocabulary, in limited replylen scope
Interior, the crucial word string of hit, the drop power for carrying out correlation are processed;
For the tendentiousness feature of thanking you, by obtaining the tendentiousness dictionary that characterizes evaluation information and by answer content and institute
State tendentiousness dictionary to be mated, carry out corresponding weighting process.
9. answer evaluation system according to claim 8, wherein, the answer characteristic information of itself includes answer content
Length and answer content paragraph number.
10. answer evaluation system according to claim 9, wherein, adjusts power mechanism determined by the weight adjustment unit
Specifically include:
For the length of answer content, corresponding tune weight coefficient is determined using the linear weighted function of multiple gears;
For the paragraph number of answer content, using the tune weight coefficient corresponding to linearly increasing form calculus.
11. answer evaluation systems according to claim 8, wherein, the user characteristics letter for submitting the answer content to
Breath includes that user gradation and user adopt rate.
12. answer evaluation systems according to claim 11, wherein, the tune power mechanism is specifically included:
Rate is adopted for user gradation and user, the tune weight coefficient being respectively adopted corresponding to corresponding Logarithmic calculation form calculus.
The 13. answer evaluation systems according to any one of claim 8-12, wherein, the tune power mechanism also includes:
If the weight for answering the characteristic information of itself is too low, drop power;
If submitting to the weight of the user's characteristic information of the answer content too low, drop power;
If it is to recommend answer, best answers etc. to answer, weighting;
If vocabulary of the short answer content comprising special marking or phrase, weighting;
Situation is answered for questioning closely, different weightings are carried out according to different ratios.
The 14. answer evaluation systems according to any one of claim 8-12, wherein, the basic weight calculation unit is led to
Cross and include regression model, based on the basic weight that each answer content is calculated to the mode that each dimension information carries out individual weighting.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310714726.7A CN103729424B (en) | 2013-12-20 | 2013-12-20 | Evaluation method and system is answered in Ask-Answer Community |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310714726.7A CN103729424B (en) | 2013-12-20 | 2013-12-20 | Evaluation method and system is answered in Ask-Answer Community |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103729424A CN103729424A (en) | 2014-04-16 |
CN103729424B true CN103729424B (en) | 2017-03-15 |
Family
ID=50453498
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310714726.7A Active CN103729424B (en) | 2013-12-20 | 2013-12-20 | Evaluation method and system is answered in Ask-Answer Community |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103729424B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106688006B (en) * | 2014-09-12 | 2020-11-20 | 艾沃瑞赛恩斯股份有限公司 | Information mediation system |
CN104462305A (en) * | 2014-11-28 | 2015-03-25 | 北京奇虎科技有限公司 | Data analyzing method and device |
CN105719159A (en) * | 2016-01-14 | 2016-06-29 | 广东小天才科技有限公司 | Method and device for motivating users |
WO2018000270A1 (en) * | 2016-06-29 | 2018-01-04 | 深圳狗尾草智能科技有限公司 | Personalized response generating method and system based on profile |
CN106776941A (en) * | 2016-12-02 | 2017-05-31 | 济南浪潮高新科技投资发展有限公司 | A kind of method of the effective solutionist of recommendation based on mass-rent pattern |
CN106802941B (en) * | 2016-12-30 | 2019-12-03 | 网易(杭州)网络有限公司 | A kind of generation method and equipment of reply message |
CN107648855A (en) * | 2017-08-29 | 2018-02-02 | 重庆邮电大学 | A kind of data processing method and device |
CN107766536A (en) * | 2017-10-30 | 2018-03-06 | 江西博瑞彤芸科技有限公司 | The searching method of related information |
CN109460823A (en) * | 2018-09-14 | 2019-03-12 | 广州神马移动信息科技有限公司 | Construction method and its device, electronic equipment, the computer-readable medium of knowledge base |
CN110032628A (en) * | 2019-02-21 | 2019-07-19 | 北京奥鹏远程教育中心有限公司 | A kind of user's on-line consulting system and method |
CN110796338A (en) * | 2019-09-24 | 2020-02-14 | 北京谦仁科技有限公司 | Online teaching monitoring method and device, server and storage medium |
CN111597313B (en) * | 2020-04-07 | 2021-03-16 | 深圳追一科技有限公司 | Question answering method, device, computer equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101441660A (en) * | 2008-12-16 | 2009-05-27 | 腾讯科技(深圳)有限公司 | Knowledge evaluating system and method in inquiry and answer community |
CN101520802A (en) * | 2009-04-13 | 2009-09-02 | 腾讯科技(深圳)有限公司 | Question-answer pair quality evaluation method and system |
CN101599164A (en) * | 2009-06-25 | 2009-12-09 | 阿里巴巴集团控股有限公司 | The method and system that the potential network client is passed judgment on |
JP4778474B2 (en) * | 2007-05-14 | 2011-09-21 | 日本電信電話株式会社 | Question answering apparatus, question answering method, question answering program, and recording medium recording the program |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8275803B2 (en) * | 2008-05-14 | 2012-09-25 | International Business Machines Corporation | System and method for providing answers to questions |
-
2013
- 2013-12-20 CN CN201310714726.7A patent/CN103729424B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4778474B2 (en) * | 2007-05-14 | 2011-09-21 | 日本電信電話株式会社 | Question answering apparatus, question answering method, question answering program, and recording medium recording the program |
CN101441660A (en) * | 2008-12-16 | 2009-05-27 | 腾讯科技(深圳)有限公司 | Knowledge evaluating system and method in inquiry and answer community |
CN101520802A (en) * | 2009-04-13 | 2009-09-02 | 腾讯科技(深圳)有限公司 | Question-answer pair quality evaluation method and system |
CN101599164A (en) * | 2009-06-25 | 2009-12-09 | 阿里巴巴集团控股有限公司 | The method and system that the potential network client is passed judgment on |
Also Published As
Publication number | Publication date |
---|---|
CN103729424A (en) | 2014-04-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103729424B (en) | Evaluation method and system is answered in Ask-Answer Community | |
EP4080889A1 (en) | Anchor information pushing method and apparatus, computer device, and storage medium | |
US7433895B2 (en) | Adding dominant media elements to search results | |
US10270791B1 (en) | Search entity transition matrix and applications of the transition matrix | |
CN104750844B (en) | Text eigenvector based on TF-IGM generates method and apparatus and file classification method and device | |
US8060456B2 (en) | Training a search result ranker with automatically-generated samples | |
CN103136329B (en) | Multiple queries revision models integrated | |
CN103593373B (en) | A kind of method and apparatus for search results ranking | |
US8332208B2 (en) | Information processing apparatus, information processing method, and program | |
CN104994424B (en) | A kind of method and apparatus for building audio and video standard data set | |
CN103064903B (en) | Picture retrieval method and device | |
CN105917364B (en) | Ranking discussion topics in question-and-answer forums | |
WO2021169347A1 (en) | Method and device for extracting text keywords | |
CN110795542A (en) | Dialogue method and related device and equipment | |
CN103699521A (en) | Text analysis method and device | |
CN110134799B (en) | BM25 algorithm-based text corpus construction and optimization method | |
CN106708940A (en) | Method and device used for processing pictures | |
CN108322317A (en) | A kind of account identification correlating method and server | |
JP2008027207A (en) | Retrieval system and retrieval method | |
JP3917629B2 (en) | Natural sentence processing apparatus and natural sentence processing program | |
CN110263973B (en) | Method and device for predicting user behavior | |
JP2001076001A (en) | Method for providing event information | |
CN101739418A (en) | Method for sequencing multi-index comprehensive weight audio-video album | |
CN113934835A (en) | Retrieval type reply dialogue method and system combining keywords and semantic understanding representation | |
CN101739417A (en) | System for sequencing multi-index comprehensive weight audio-video album |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |