CN103729424A - Method and system for assessing answers in Q&A (questions and answers) community - Google Patents
Method and system for assessing answers in Q&A (questions and answers) community Download PDFInfo
- Publication number
- CN103729424A CN103729424A CN201310714726.7A CN201310714726A CN103729424A CN 103729424 A CN103729424 A CN 103729424A CN 201310714726 A CN201310714726 A CN 201310714726A CN 103729424 A CN103729424 A CN 103729424A
- Authority
- CN
- China
- Prior art keywords
- answer
- answer content
- information
- weight
- feature
- 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
Images
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)
- Information Transfer Between Computers (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method for assessing answers in a Q&A (questions and answers) community. The method includes acquiring a question, all answer content corresponding to the question and multi-dimensional information related to the answer content; according to independent weighting on the multi-dimensional information, calculating a basic weight of each answer content; synthesizing mutual influences of the multi-dimensional information, and determining a corresponding weight adjusting mechanism to adjust the acquired basic weights to acquire a final weight of each answer content; based on the final weights, assessing all the answer content comprehensively. Correspondingly, the invention further provides a system for assessing the answers in the Q&A community. By the method and the system, answers valuable to questions can be effectively distinguished, and user experience of a Q&A platform is improved.
Description
Technical field
The present invention relates to computer network field, relate in particular to answer evaluation method and system in a kind of Ask-Answer Community.
Background technology
At present, by search platform, search for the important channel that relevant information is user's obtaining information, especially in Ask-Answer Community search problem, ask a question, answer a question, browse problem or append problem etc., this has become the important way of carrying out interactive information interchange between user.Wherein, common Ask-Answer Community have Baidu know, search ask, Sina likes to ask etc.
Conventionally, in Ask-Answer Community under each problem the displaying of answer content order mainly based on following two kinds of modes: 1) only according to the time of answering a question, sort, i.e. the answer of the forward displaying of rank is the time of more close current search in time; 2) according to answering the favorable comment number obtaining, sort, i.e. the approval number that answer under same problem obtains user is more, its more forward Ask-Answer Community that is illustrated in.But, these two kinds of modes respectively have its deficiency, for first kind of way, because the answer of this problem is not necessarily mated most in the answer of forward displaying, therefore, user conventionally need to take a long time and find needed answer, and, this mode is along with answering increasing progressively of number, and its deficiency is more obvious; For the second way, based on approval is several, to answering, sort, this is easy to suffer spam(electronic waste) user's attack, make those to spam user directly useful ad content push up forward display location, thereby cause browsing the user's of this answer misleading.
Summary of the invention
The object of this invention is to provide in a kind of Ask-Answer Community and answer evaluation method and system, the user that can effectively promote answer platform experiences.
According to an aspect of the present invention, provide in a kind of Ask-Answer Community and answered evaluation method, the method comprises:
Obtain all answer content corresponding under problem and described problem and the multidimensional information relevant to described answer content;
By including regression model in, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content;
Comprehensive respectively tie up influencing each other of information, determine the corresponding basic weight of adjusting power mechanism to obtain described in regulating, obtain the final weight of each answer content;
Based on described final weight, described all answer content are carried out to comprehensive evaluation.
According to another aspect of the present invention, also provide in a kind of Ask-Answer Community and answered evaluation system, having comprised:
Information acquisition unit, for obtaining all answer content corresponding under problem and described problem and the multidimensional information relevant to described answer content;
Basic weight calculation unit, by including regression model in, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content;
Weight regulon, for comprehensively respectively tieing up influencing each other of information, determines the corresponding basic weight of adjusting power mechanism to obtain described in regulating, obtains the final weight of each answer content;
Answer evaluation unit, based on described final weight, described all answer content are carried out to comprehensive evaluation.
Compared with prior art, the present invention has the following advantages:
1) the present invention, by answering the assessment of information, effectively screens the valuable answer of problem, and this answer is preferentially represented to viewer and quizmaster, and the user who has promoted answer platform experiences;
2) the present invention can prevent junk information (spam) user's attack effectively, avoids this category information to cause misleading to browsing user.
Accompanying drawing explanation
By reading the detailed description that non-limiting example is done of doing with reference to the following drawings, it is more obvious that other features, objects and advantages of the present invention will become:
Fig. 1 answers evaluation method process flow diagram in Ask-Answer Community in accordance with a preferred embodiment of the present invention;
Fig. 2 is the length of the answer content shown in the present embodiment and the curve map of corresponding tune weight coefficient;
Fig. 3 is the graph of relation of the quality of user gradation and answer content according to the preferred embodiment of the invention;
Fig. 4 answers the schematic block diagram of evaluation system in the answer community of another preferred embodiment of the present invention.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in further detail.
According to an aspect of the present invention, provide answer evaluation method in a kind of Ask-Answer Community.It should be noted that, the weight of below mentioning becomes positive relationship with the quality of the information of answer, and weight is higher, represents that the quality of answer information is more excellent.The quality of described answer information mainly from answer content, submit to the user behavior feature of described answer, the information such as user characteristics of browsing described answer comprehensively to weigh.
Please refer to Fig. 1, Fig. 1 answers evaluation method process flow diagram in Ask-Answer Community in accordance with a preferred embodiment of the present invention.
As shown in Figure 1, method provided by the present invention comprises the following steps:
Step S101, obtains all answer content corresponding under problem and described problem and the multidimensional information relevant to described answer content.
Particularly, in order better the value of the answer information in Ask-Answer Community to be evaluated, Network Basedly obtain in Ask-Answer Community all answer content and relevant information corresponding under all problems and described problem, this is not restricted for the mode of specifically obtaining.
The multidimensional information relevant to described answer content mainly comprises: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.Wherein, the characteristic information of described answer self comprises non-text characteristics information and text feature information; The user's characteristic information of the described answer content of described submission comprises that user gradation and user adopt rate; The described user behavior feature of browsing described problem and answer mainly refers to that this user is to the evaluation information of answering, the language and question closely the language etc. of thanking you in answering of thanking you such as common answer favorable comment number, in answering, this information can be portrayed the feedback information of this user to this answer.
Wherein, the text feature information spinner in the characteristic information of described answer self will comprise: special marking feature, core presentive word feature, query tendency feature and meaningless feature, the tendentiousness of thanking you feature.
Wherein, the non-text feature information spinner in the characteristic information of described answer self will comprise: the paragraph number of the length information of described answer content, described answer content, Rich Media's characteristic information are or/and question closely the information of answering.Wherein, Rich Media's characteristic information mainly refers to characteristic informations such as picture, map in answer content.
Step S102, by including regression model in, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content.
Particularly, obtain after above-mentioned multidimensional information, based on above-mentioned multidimensional information, weigh the quality of each answer content.More specifically, by following computing formula, calculate the basic weight of each answer content, include regression model in, described each dimension information is carried out to linear weighted function calculating.Computing formula is as follows:
score
ini=radio
1×dimesion
1+…radio
i×dimesion
i+…radio
n×dimesion
n
Wherein, radio
1, radio
i, radio
nrepresent respectively the tune weight factor of Ge Wei information, dimesion
1, dimesion
i, dimesion
nrepresent respectively the weight of Ge Wei information, score
inirepresent basic weight.The methods such as exhaustive, the selection that wherein, described tune weight factor and weight can be by the feature to included in each dimension information and main characteristic informations, cure parameter are determined.
The described information spinner of respectively tieing up will refer to: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.
By the calculating of above-mentioned formula, can obtain the basic weight of each answer content, by tune power below, process, can obtain the final weight of each answer content.
Step S103, comprehensively respectively ties up influencing each other of information, determines the corresponding basic weight of adjusting power mechanism to obtain described in regulating, obtains the final weight of each answer content.
Particularly, described mainly each dimension information of finger that influences each other of respectively tieing up information produces positive or negative impact to the quality of answer content.Described tune power mechanism according to described impact just negative, degree of influence is weighted or fall power, in the basic weight of each answer content, determines the corresponding weight coefficient of adjusting, and comprises weighting or falls weight coefficient, both products are the weight of final acquisition.Concrete computing formula can be with reference to below:
score=w
1×…W
m×score
ini
Wherein, w
1, w
mrepresent to adjust weight coefficient, score
inirepresent basic weight, score represents final weight.
Wherein, described weighting or to fall weight coefficient relevant to the specific features in each dimension information, below will describe in detail.
With regard to the non-text feature information in the characteristic information of above-mentioned answer self, the feature that wherein affects weighting or fall weight coefficient mainly comprises the paragraph of length and the answer content of answer content.
Respectively, because the length of answer content in Ask-Answer Community is all generally that its quality of content of moderate-length is higher, the information that the content of too short length comprises is conventionally more unilateral, its quality is lower, long content is conventionally because tediously long and lack keynote message, therefore, the length of answer content presents and first increases the trend reducing afterwards the contribution of answer content quality.In order to embody better the relation between length and answer content quality, described relation curve can be divided into a plurality of gears and represent, and can adopt following formula to calculate the corresponding tune weight coefficient of length of described answer content:
Wherein, len represents the supposition length of answer content, and 1 to n represents respectively to adjust weight coefficient w
1to w
ncorresponding gear, len
1to len
nrepresent respectively 1 length of interval to n shelves correspondence, w
lengththe corresponding tune weight coefficient of length of the described answer content that expression finally obtains.
Further, can be with reference to figure 2, Fig. 2 is the length of the answer content shown in the present embodiment and the curve map of corresponding tune weight coefficient.As shown in Figure 2, the calculating of the tune weight coefficient of described answer content length adopts the weighted calculation mode of above-mentioned a plurality of gears, tune weight coefficient between adjacent gear adopts the tune weight coefficient of adjacent low-grade location and the phase Calais of variable to obtain, and the tune weight coefficient finally obtaining and the corresponding relation of content-length can adopt as (len
1, w
1), (len
2, w
2), (len
3, w
3) ... (len
n, w
n) etc. with the form of (length, adjust weight coefficient), represent.
Further, as described above, the paragraph number of answer content is also closely related with tune weight coefficient, and it can embody the structurized fine or not degree of answer text, can adopt particularly the linear form increasing to calculate the tune weight coefficient corresponding to paragraph of varying number, can be with reference to following formula:
Wherein, p
radiothat paragraph is adjusted power radix; p
numit is current answer paragraph number; p
topthat the answer paragraph of setting is counted threshold value, w
paragraphfor tune weight coefficient corresponding to described paragraph calculating.
Further, for the Rich Media's characteristic information in described non-text feature information, answer content comprises characteristic informations such as picture, map, directly in the basic weight of described answer content, is weighted.Equally, for comprising the answer content of questioning closely the information of answering, according to described length information, paragraph number and the Rich Media's characteristic information etc. of answering content of questioning closely, carry out corresponding tune power and process.
With regard to submitting the user's characteristic information of described answer content to, the feature that wherein affects weighting or fall weight coefficient mainly comprises that user gradation and user adopt rate.
Discuss respectively, conventionally user gradation is higher, the possibility that this answer content quality is high is higher, but to a certain extent, gradually mild, can be with reference to figure 3, Fig. 3 is the graph of relation of the quality of user gradation and answer content according to the preferred embodiment of the invention, as shown in Figure 3, the quality of answer content, along with the growth of user gradation presents the mild progressive variation tendency of first increasing sharply again, can adopt following formula (being the form of Logarithmic calculation) to calculate the corresponding tune weight coefficient of user gradation:
Wherein, level
radiorepresent user gradation weighting factor, u
levelthe grade that represents this user, top
levelthe highest user gradation of setting, w
user gradationrepresent the final tune weight coefficient that described user gradation is corresponding.
Further, it is the probability that user's answer is adopted that user adopts rate, it can weigh the quality of the historical answer content of this user, situation about being adopted according to the historical answer content of this user, can predict that this user contributes the possibility of high-quality answer, to this, can adopt following formula (being the form of Logarithmic calculation) to calculate described user and adopt tune weight coefficient corresponding to rate:
Wherein, good_rate
radiorepresentative of consumer is adopted rate weighting factor; Good_rate represents that this user adopts rate; top
good_ratethat the highest user who sets adopts rate value, w
adopt raterepresent that described user adopts final tune weight coefficient corresponding to rate.
With regard to browsing the user behavior characteristic information of described problem and answer, the feature that wherein affects weighting or fall weight coefficient mainly comprises answers favorable comment number, the user tendentiousness feature of language of thanking you, and questions closely the tendentiousness feature etc. of thanking you of answering the inside.
Wherein, answer the feature of favorable comment number as user behavior characteristic information, for portraying user to the feedback information of answering, the good evaluating data of major embodiment user to certain answer of seeing, relation object between this feature and tune weight coefficient is similar to above-mentioned user to be adopted rate and adjusts the corresponding relation between weight coefficient, can adopt equally the mode of Logarithmic calculation, for simplicity's sake, at this, be not repeated.
With regard to the text feature information in the characteristic information of described answer self, the feature that wherein affects weighting or fall weight coefficient mainly comprises special marking feature, core presentive word feature, query tendency feature and meaningless feature, the tendentiousness of thanking you feature.Below putting up with these four features describes in detail.
For answer content, comprise special marking features such as < < > >, " ", <>, carry out respective weight processing.
Wherein, described core presentive word refers to the core word that reverse file frequency weight surpasses certain threshold value and filters through stop words, symbol, short number word alphabetic string etc.In the present embodiment, to described core presentive word, the analysis in the power of tune mechanism mainly comprises two steps: 1) generate core vocabulary; 2) coupling core word.
Particularly, with regard to step 1), main by the word frequency information in statistical problem title and filter (such as filtering stop words wherein, symbol, short number word alphabetic string etc.), calculate the reverse file frequency of idf(of word, inverse document frequency) distribute and form the vocabulary that comprises reverse file frequency weight information.
With regard to step 2) with regard to, be mainly divided into following several step and carry out:
I) set certain threshold value, extract weight in described problem title and be greater than the word of described threshold value and sort by weight, retain forward maximum 2 core words (referred to as word 1, word 2) of rank;
II), in the core vocabulary of described formation, expand institute's predicate 1 or/and the synonym of word 2;
III) adjust the weight of institute's predicate 1, word 2, if the idf weight of two words differs larger, power is fallen in word 2 and process, to give prominence to the impact of the keyword that competency is strong;
IV) fetch and answer maximum top n bytes and mate with the core word of described reservation, and the idf weights stepping of coupling is smoothly mapped between designation area upper, to avoid the noise of the rear section content of long answer to exert an influence to the coupling of core word.
Wherein, described query tendency feature and meaningless feature refer to and in answer content, include query tendency or the insignificant situation of content itself.Conventionally, in answer content, why being with the tendency that has a question is because problem is unclear.Give an example:
(1) problem: what if invalid trade mark registration is?
Answer: this must see that your trade mark is that what reason is invalid.
(2) problem: how much this needs what specifically needs across web game to act on behalf of angle road
Answer: what agency online friend needs
(3) problem: why excel that may I ask me has become two kinds of icons below, has changed unfolding mode and has not also used
Answer: do not meet such icon, too shy, can't help busy.
By above-mentioned three examples, the answer content in (1) and (2) belongs to the situation that includes query tendency, and the answer content in (3) belongs to the insignificant situation of content itself.
For the answer that comprises described query tendency feature and meaningless feature, the form of mainly mating by vocabulary, within the scope of limited replylen, hits keyword string, falls accordingly power and processes.
Wherein, the tendentiousness of thanking you described in feature comprises the evaluation information of forward, negative sense or the other types of user in answer.When analyzing this feature and adjusting being related between weight coefficient, first, the words and phrases frequency of thanking you in answering by statistics, and by obtain the tendentiousness dictionary of the positive and negative evaluation information of obvious sign such as the mode of manually checking (review); Secondly, coupling tendentiousness dictionary, carries out tendentiousness judgement according to forward evaluation and optimization in the principle of negative sense evaluation, if the result of judging is as hitting forward vocabulary, is weighted; Hit negative sense vocabulary, be not weighted; Otherwise, for this answer, do not comprise the situation of described tendentiousness dictinary information, other situations based on mentioned above are carried out respective weight processing.
Generally speaking, on the basis based on mentioned above, the weighting in the present embodiment or fall power mechanism and also comprise following situation:
The weight of the characteristic information of described answer self is too low, falls power;
Submit to the weight of user's characteristic information of described answer content too low, fall power;
Answer is to recommend answer, best answers etc., weighting;
The vocabulary that short answer comprises special marking or phrase, weighting;
For questioning closely the situation of answering, according to different ratios, carry out different weightings.
Based on the above-mentioned basic weighting of enumerating and the mode of falling power, on the basis of the original answer weight of calculating, to answering, carry out corresponding weighting and fall power and process, generate the final weight of described answer content.
Step S104, carries out comprehensive evaluation based on described final weight to described all answer content.
Particularly, according to described final weight, described all answer content are sorted, the forward answer content of rank is evaluated as best answers, the answer content ranking behind is evaluated as suboptimum and is answered, and preferably according to sequence, shows from high to low described answer content and relevant information on the page.
Compared with prior art, method provided by the present invention has the following advantages: answer according to the method for the value auto-sequencing of puing question to is made preferentially to represent the valuable answer to problem and become possibility, the method can be optimized the sortord of answering under millions quantity problem, making to browse user priority sees the more helpful answer of dealing with problems, thereby minimizing is browsed user and is arrived the time satisfying the demands after the page and searching energy cost, optimize viewing experience, promote and browse satisfaction.
According to another aspect of the present invention, also provide in a kind of Ask-Answer Community and answered evaluation system, please refer to Fig. 4, in the answer community that Fig. 4 is another preferred embodiment of the present invention, answered the schematic block diagram of evaluation system.As shown in Figure 4, this system comprises:
Basic weight calculation unit 402, by including regression model in, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content;
Answer evaluation unit 404, based on described final weight, described all answer content are carried out to comprehensive evaluation.
Below, the course of work of each unit provided by the present invention is specifically described.
Particularly, in order better the value of the answer information in Ask-Answer Community to be evaluated, described information acquisition unit 401 is Network Based obtains in Ask-Answer Community corresponding all answer content and relevant information under all problems and described problem, and this is not restricted for the mode of specifically obtaining.The multidimensional information relevant to described answer content mainly comprises: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.Wherein, the characteristic information of described answer self comprises non-text characteristics information and text feature information; The user's characteristic information of the described answer content of described submission comprises that user gradation and user adopt rate; The described user behavior feature of browsing described problem and answer mainly refers to that this user is to the evaluation information of answering, the language and question closely the language etc. of thanking you in answering of thanking you such as common answer favorable comment number, in answering, this information can be portrayed the feedback information of this user to this answer.
Wherein, the text feature information spinner in the characteristic information of described answer self will comprise: special marking feature, core presentive word feature, query tendency feature and meaningless feature, the tendentiousness of thanking you feature.
Wherein, the non-text feature information spinner in the characteristic information of described answer self will comprise: the paragraph number of the length information of described answer content, described answer content, Rich Media's characteristic information are or/and question closely the information of answering.Wherein, Rich Media's characteristic information mainly refers to characteristic informations such as picture, map in answer content.
Obtain after above-mentioned multidimensional information, based on above-mentioned multidimensional information, weigh the quality of each answer content, and by basic weight calculation unit 402, by following computing formula, calculated the basic weight of each answer content, and include regression model in, described each dimension information is carried out to linear weighted function calculating.Computing formula is as follows:
score
ini=radio
1×dimesion
1+…radio
i×dimesion
i+…radio
n×dimesion
n
Wherein, radio
1, radio
i, radio
nrepresent respectively the tune weight factor of Ge Wei information, dimesion
1, dimesion
i, dimesion
nrepresent respectively the weight of Ge Wei information, score
inirepresent basic weight.Wherein respectively tieing up information spinner will refer to: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.By the calculating of above-mentioned formula, can obtain the basic weight of each answer content.
Wherein, comprehensive each dimension information of described weight regulon 403 produces positive or negative impact to the quality of answer content.Determine corresponding weighting or fall the power mechanism basic weight to obtain described in regulating, particularly, determine corresponding weighting or fall weight coefficient in the basic weight of each answer content, both products are the weight of final acquisition.Concrete computing formula can be with reference to below:
score=w
1×…w
m×score
ini
Wherein, w
1, w
mrepresent to adjust weight coefficient, score
inirepresent basic weight, score represents final weight.Wherein, described weighting or to fall weight coefficient relevant to the specific features in each dimension information.Due to each feature and adjust relation between weight coefficient as described above, for simplicity's sake, no longer describe in detail.
Wherein, described answer evaluation unit 404 sorts and evaluates described all answer content according to described final weight, and the answer content that rank is forward is evaluated as best answers, and the answer content ranking behind is evaluated as suboptimum and answers.
Preferably, this system also comprises display unit, for according to the sequence of final weight, shows from high to low described answer content and relevant information on the page.
This system provided by the present invention has the following advantages: native system is by the processing of basic weight calculation unit and weight regulon, can pick out preferably the answer that problem is had to higher-value, and can promote the experience of answer platform for user according to the sequence being worth.
Above disclosed is only preferred embodiment of the present invention, certainly can not limit with this interest field of the present invention, and the equivalent variations of therefore doing according to the claims in the present invention, still belongs to the scope that the present invention is contained.
Claims (20)
1. the answer evaluation method in Ask-Answer Community, the method comprises:
A) obtain all answer content corresponding under problem and described problem and the multidimensional information relevant to described answer content;
B) mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content;
C) comprehensively respectively tie up influencing each other of information, determine the corresponding basic weight of adjusting power mechanism to obtain described in regulating, obtain the final weight of each answer content;
D) based on described final weight, described all answer content are carried out to comprehensive evaluation.
2. answer evaluation method according to claim 1, wherein, described multidimensional information mainly comprises: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.
3. answer evaluation method according to claim 2, wherein, the characteristic information of described answer self comprises the paragraph number of length and the answer content of answer content.
4. answer evaluation method according to claim 3, wherein, described tune power mechanism specifically comprises:
For the length of answer content, adopt the linear weighted function of a plurality of gears to determine corresponding tune weight coefficient;
For the paragraph number of answer content, adopt the linear form increasing to calculate corresponding tune weight coefficient.
5. answer evaluation method according to claim 2, wherein, the user's characteristic information of the described answer content of described submission comprises that user gradation and user adopt rate.
6. answer evaluation method according to claim 5, wherein, described tune power mechanism specifically comprises:
For user gradation and user, adopt rate, adopt respectively corresponding Logarithmic calculation form to calculate corresponding tune weight coefficient.
7. answer evaluation method according to claim 2, wherein, the characteristic information of described answer self comprises special marking feature, core presentive word feature, query tendency feature and meaningless feature, the tendentiousness of thanking you feature.
8. answer evaluation method according to claim 7, wherein, described tune power mechanism specifically comprises:
For described special marking feature, be directly weighted processing;
For described core presentive word feature, by generating core vocabulary and mating core word, determine corresponding tune weight coefficient;
For described query tendency feature and meaningless feature, the form of mating by vocabulary, within the scope of limited replylen, hits keyword string, and the power of falling of being correlated with is processed;
For the described tendentiousness feature of thanking you, by obtaining, characterize the tendentiousness dictionary of evaluation information and answer content is mated with described tendentiousness dictionary, carry out corresponding weighting processing.
9. according to the answer evaluation method described in claim 2-8 any one, wherein, described tune power mechanism also comprises:
If the weight of the characteristic information of described answer self is too low, power is fallen;
If submit to the weight of user's characteristic information of described answer content too low, fall power;
If answer, be to recommend answer, best answers etc., weighting;
If the vocabulary that short answer content comprises special marking or phrase, weighting;
For questioning closely the situation of answering, according to different ratios, carry out different weightings.
10. according to the answer evaluation method described in claim 2-8 any one, wherein, described step b) specifically comprises:
By including regression model in, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content.
Answer evaluation system in 11. 1 kinds of Ask-Answer Communities, comprising:
Information acquisition unit, for obtaining all answer content corresponding under problem and described problem and the multidimensional information relevant to described answer content;
Basic weight calculation unit, the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content;
Weight regulon, for comprehensively respectively tieing up influencing each other of information, determines the corresponding basic weight of adjusting power mechanism to obtain described in regulating, obtains the final weight of each answer content;
Answer evaluation unit, based on described final weight, described all answer content are carried out to comprehensive evaluation.
12. answer evaluation systems according to claim 11, wherein, described multidimensional information mainly comprises: the characteristic information of described answer self, submit to described answer content user's characteristic information, browse the user behavior characteristic information of described problem and answer.
13. answer evaluation systems according to claim 12, wherein, the characteristic information of described answer self comprises the paragraph number of length and the answer content of answer content.
14. answer evaluation systems according to claim 13, wherein, the determined tune power of described weight regulon mechanism specifically comprises:
For the length of answer content, adopt the linear weighted function of a plurality of gears to determine corresponding tune weight coefficient;
For the paragraph number of answer content, adopt the linear form increasing to calculate corresponding tune weight coefficient.
15. answer evaluation systems according to claim 12, wherein, the user's characteristic information of the described answer content of described submission comprises that user gradation and user adopt rate.
16. answer evaluation systems according to claim 15, wherein, described tune power mechanism specifically comprises:
For user gradation and user, adopt rate, adopt respectively corresponding Logarithmic calculation form to calculate corresponding tune weight coefficient.
17. answer evaluation systems according to claim 12, wherein, the characteristic information of described answer self comprises special marking feature, core presentive word feature, query tendency feature and meaningless feature, the tendentiousness of thanking you feature.
18. answer evaluation systems according to claim 17, wherein, described tune power mechanism specifically comprises:
For described special marking feature, be directly weighted processing;
For described core presentive word feature, by generating core vocabulary and mating core word, determine corresponding tune weight coefficient;
For described query tendency feature and meaningless feature, the form of mating by vocabulary, within the scope of limited replylen, hits keyword string, and the power of falling of being correlated with is processed;
For the described tendentiousness feature of thanking you, by obtaining, characterize the tendentiousness dictionary of evaluation information and answer content is mated with described tendentiousness dictionary, carry out corresponding weighting processing.
19. according to the answer evaluation system described in claim 12-18 any one, and wherein, described tune power mechanism also comprises:
If the weight of the characteristic information of described answer self is too low, power is fallen;
If submit to the weight of user's characteristic information of described answer content too low, fall power;
If answer, be to recommend answer, best answers etc., weighting;
If the vocabulary that short answer content comprises special marking or phrase, weighting;
For questioning closely the situation of answering, according to different ratios, carry out different weightings.
20. according to the answer evaluation system described in claim 12-18 any one, and wherein, described basic weight calculation unit is by including regression model in, and the mode based on each dimension information is carried out to independent weighting is calculated the basic weight of each answer content.
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 true CN103729424A (en) | 2014-04-16 |
CN103729424B 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) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN106688006A (en) * | 2014-09-12 | 2017-05-17 | 艾沃瑞赛恩斯股份有限公司 | Information brokering system |
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 |
CN106802941A (en) * | 2016-12-30 | 2017-06-06 | 网易(杭州)网络有限公司 | The generation method and equipment of a kind of reply message |
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 |
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 |
CN111597313A (en) * | 2020-04-07 | 2020-08-28 | 深圳追一科技有限公司 | Question answering method, device, computer equipment and storage medium |
Citations (5)
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 |
US20090287678A1 (en) * | 2008-05-14 | 2009-11-19 | International Business Machines Corporation | System and method for providing answers to questions |
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 |
-
2013
- 2013-12-20 CN CN201310714726.7A patent/CN103729424B/en active Active
Patent Citations (5)
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 |
US20090287678A1 (en) * | 2008-05-14 | 2009-11-19 | International Business Machines Corporation | System and method for providing answers to questions |
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 |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106688006A (en) * | 2014-09-12 | 2017-05-17 | 艾沃瑞赛恩斯股份有限公司 | Information brokering 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 |
CN106802941A (en) * | 2016-12-30 | 2017-06-06 | 网易(杭州)网络有限公司 | The generation method and equipment of a kind 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 |
CN111597313A (en) * | 2020-04-07 | 2020-08-28 | 深圳追一科技有限公司 | Question answering method, device, computer equipment 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 |
Also Published As
Publication number | Publication date |
---|---|
CN103729424B (en) | 2017-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103729424A (en) | Method and system for assessing answers in Q&A (questions and answers) community | |
JP5731250B2 (en) | System and method for recommending interesting content in an information stream | |
US8290927B2 (en) | Method and apparatus for rating user generated content in search results | |
CN103425635B (en) | Method and apparatus are recommended in a kind of answer | |
CN103399891B (en) | Method for automatic recommendation of network content, device and system | |
US7711735B2 (en) | User segment suggestion for online advertising | |
EP1818839A1 (en) | System and method for online information analysis | |
US8332208B2 (en) | Information processing apparatus, information processing method, and program | |
CN105095187A (en) | Search intention identification method and device | |
CN105917364B (en) | Ranking discussion topics in question-and-answer forums | |
CN108304512A (en) | A kind of thick sort method of video search engine, device and electronic equipment | |
CN106547871A (en) | Method and apparatus is recalled based on the Search Results of neutral net | |
CN103699521A (en) | Text analysis method and device | |
CN111026868B (en) | Multi-dimensional public opinion crisis prediction method, terminal device and storage medium | |
CN106649761A (en) | Search result display method and device based on profound questioning and answering | |
CN110532450B (en) | Topic crawler method based on improved shark search | |
CN110737859A (en) | UP main matching method and device | |
CN112579729B (en) | Training method and device for document quality evaluation model, electronic equipment and medium | |
CN110175264A (en) | Construction method, server and the computer readable storage medium of video user portrait | |
CN108717469B (en) | Post sorting method, device and equipment and computer readable storage medium | |
CN110275943A (en) | Article method for pushing and device | |
CN110609950B (en) | Public opinion system search word recommendation method and system | |
US8838616B2 (en) | Server device for creating list of general words to be excluded from search result | |
CN112612961A (en) | Information searching method and device, storage medium and computer equipment | |
CN112184021A (en) | Answer quality evaluation method based on similar support set |
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 |