CN101169797B - Searching method - Google Patents

Searching method Download PDF

Info

Publication number
CN101169797B
CN101169797B CN2007101882729A CN200710188272A CN101169797B CN 101169797 B CN101169797 B CN 101169797B CN 2007101882729 A CN2007101882729 A CN 2007101882729A CN 200710188272 A CN200710188272 A CN 200710188272A CN 101169797 B CN101169797 B CN 101169797B
Authority
CN
China
Prior art keywords
document
degree
correlation
search results
search
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.)
Expired - Fee Related
Application number
CN2007101882729A
Other languages
Chinese (zh)
Other versions
CN101169797A (en
Inventor
朱廷劭
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Individual
Original Assignee
Individual
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Individual filed Critical Individual
Priority to CN2007101882729A priority Critical patent/CN101169797B/en
Publication of CN101169797A publication Critical patent/CN101169797A/en
Application granted granted Critical
Publication of CN101169797B publication Critical patent/CN101169797B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a method for optimizing search results, which comprises acquiring correlation evaluation of files containing search result; and sequencing the files containing search result according to the correlation evaluation of the files containing search result and content of the files containing search result. The inventive method for optimizing search result can make clear the actual search intent of user according to the correlation evaluation of files containing search result fed back from user, can sequence the files containing search result in combination with the files containing search result, and can feed files actually interested by user, useful to user and meeting user's search intent back to the user with high priority, so as to remarkably reduce user's search time, and improve search efficiency. By using the invention, users can rapidly, accurately and conveniently acquire the information required by them.

Description

A kind of method that is used to search for
Technical field
The present invention relates to search technique, be meant a kind of method that is used to search for especially.
Background technology
Constantly grown up since being born in the internet, its content is enriched constantly, and whole network is piled into a unprecedented ultra-large type information bank gradually.The internet is being brought into play more and more important effect as an information platform in daily life and work, people obtain information by the internet more and more.Follow the volatile development in internet, the network user wants to find required information simply as looking for a needle in a haystack, and is at a loss to such an extent as to get lost in the ocean of information.Obtaining the information that oneself needs how fast, accurately, easily from so huge information bank, is the major issue that the Internet user faces.
The processing procedure of search engine is at present, after the user imports one or more search keys, searches for according to the search key of user's input, searches out the document that comprises these key words and returns to the user as Search Results.
In the search of reality, both having made is identical searching key word, because the search intention difference of different user, different users wants that really the content of looking for may be different fully.Because in present search engine, Search Results is not optimized and returns to the user, therefore, might not be that this user is real to the document that the user returned wants.Like this, the user can not obtain needed information fast, accurately, easily.
Summary of the invention
In view of this, fundamental purpose of the present invention is to provide a kind of method that is used to search for, and returns the document that meets user search intent for the user.
In order to achieve the above object, the invention provides a kind of method that is used to search for, this method comprises:
Obtain correlativity evaluation for the document of Search Results;
Obtain correlativity evaluation for the document of Search Results;
According to the correlativity evaluation of the document of Search Results and the document content of Search Results, the document of Search Results is sorted, specifically comprise:
According to the document content of Search Results, calculate the document similarity between any two of Search Results;
Obtain the similarity relation figure of the document of Search Results, the node of described similarity relation figure is the document of Search Results, and the weight between the node of described similarity relation figure is the similarity between the corresponding document of described node;
According to the correlativity evaluation of the document of Search Results and the document similarity relation figure of Search Results, the degree of correlation of each document of initialization, for obtaining the document that correlativity is estimated, the initial value of its degree of correlation is to estimate corresponding value with correlativity, for the document that does not have correlativity to estimate, the initial value of its degree of correlation is a default value;
The document that utilizes Search Results between any two similarity and the initial value of the degree of correlation or last cyclic process in the degree of correlation that obtains, upgrade the degree of correlation that obtains each document in this cyclic process;
Judge whether to reach the cycle index of regulation, whether the degree of correlation of judging each document is stable, if reaching the cycle index of regulation or the degree of correlation of each document stablizes, then with the degree of correlation of each document of obtaining in this circulation the degree of correlation, otherwise return the step of carrying out the degree of correlation of upgrading each document as each document;
Sort according to the degree of correlation of resulting each document document to Search Results.
Preferably, the step of described calculating document similarity between any two comprises: obtain the speech that each document comprises, obtain the number of the speech that each document comprises; When first document in calculating search result document and the similarity between second document, obtain the number of the speech that described first document and second document comprise jointly; The number of the speech that is comprised with described first document and the number sum of the speech that second document is comprised deduct the value that the number of the speech that described first document and second document comprise jointly obtains again, remove the resulting value of number of the speech that described first document and second document comprise jointly, as the similarity between described first document and second document.
Preferably, describedly judge whether the degree of correlation of each document stabilizes to: whether square sum of difference of judging the degree of correlation that obtains in the degree of correlation that obtains in this circulation and the last circulation is less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise be judged to be stable; Perhaps, judge the degree of correlation that obtains in the degree of correlation that obtains and last circulation in this circulation difference square root sum square whether less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise be judged to be stable.
Preferably, when the circulation time first time, the degree of correlation of each document is in described this cyclic process, w i T = Σ j = 1 N ( w j 0 * J ( d j , d i ) ) , Wherein, w j 0Document d jThe initial value of the degree of correlation, J (d j, d i) be document d jWith document d iBetween similarity, w i TBe the T time document d in the circulation iDegree of correlation score, N is the number of documents of Search Results, T be the T time the circulation;
When the above circulation time second time, the degree of correlation of each document is in described this cyclic process, w i T = Σ j = 1 N ( w j T - 1 * J ( d j , d i ) ) , Wherein, w j T-1Be the document d that in last once circulation, obtains jThe degree of correlation, w i TBe the degree of correlation score of the T time document di in the circulation, J (d j, d i) be document d jWith document d iBetween similarity, N is the number of documents of Search Results, T be the T time the circulation.
Preferably, described renewal obtains after the degree of correlation of each document in this cyclic process, before carrying out described judgement, further comprise: search maximal value in the degree of correlation of each document that from this circulation, obtains, and according to this maximal value the degree of correlation of each document is carried out normalizing and handle.
The correlativity evaluation of described document for Search Results is selected one of them option feedback by the user from a plurality of evaluation options that search system provides, perhaps feed back by the mode of answering the problem that search system proposed, perhaps the mode that directly provides evaluation score by the user is fed back.
The Search Results optimization method that provides by the present invention, come clear and definite user's real search intention according to user feedback for the correlativity evaluation of search result document, and in conjunction with the document content of Search Results, document to Search Results is optimized ordering, the user is really interested, the useful document that meets user search intent of user is returned to the user with high priority, thereby significantly reduce user's search time, make the user can obtain self needed information fast, accurately, easily, improved search efficiency.The present invention is further according to the similarity between the search result document content calculating document, and utilize the degree of correlation of each document of the correlativity evaluation calculation Search Results of similarity between the document and user feedback, thereby avoid the user to participate in search procedure too much, under the situation that a spot of correlativity of user feedback is estimated, search system also can obtain by the degree of correlation of calculating each document automatically and user search intent between degree of correlation, return the document that meets search intention for the user.
Description of drawings
Figure 1 shows that the synoptic diagram of in the embodiment of the invention one Search Results being optimized;
Figure 2 shows that the process flow diagram that calculates the file correlation score in the embodiment of the invention one;
Figure 3 shows that the synoptic diagram of in the embodiment of the invention two Search Results being optimized.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, lift specific embodiment below, the present invention is further detailed explanation.
In the method to Search Results optimization provided by the invention, after obtaining Search Results by search key, according to the document content of Search Results and to the correlativity evaluation of the document of Search Results, Search Results is sorted, will return to the user with relative high priority with the document that user search intent is consistent.
When in the present invention Search Results being optimized, mainly consider two factors, the one, to the correlativity evaluation of the document of Search Results feedback, the 2nd, the content of the document of Search Results.Consider the correlativity evaluation of user to the document feedback of Search Results, the user's that then can sharpen understanding search intention can return the document that meets user search intent for the user so better.Obtain the correlativity evaluation that document fed back to Search Results, and consider the document content of Search Results again, the degree of correlation that then can know each document of Search Results and user search intent how.Like this, under the situation that the correlativity of a small amount of document of user feedback is estimated, can access each document of Search Results and the degree of correlation of user search intent, promptly be equivalent to obtain user's correlativity evaluation, thereby can return the document that meets user search intent for the user.
Embodiment one:
Fig. 1 is the synoptic diagram of among the embodiment one Search Results being optimized.As shown in Figure 1, the process to Search Results optimization may further comprise the steps:
Step 101, the key word of importing according to the user obtains Search Results.
Comprise the document of type of webpage and the document that can obtain its content of other types in the Search Results.Suppose to comprise in the Search Results N document, use d respectively 1, d 2..., d i..., d j..., d NExpression.
In addition, the key word of user's input can be one, also can be a plurality of.
Step 102 for the document of Search Results, is calculated document similarity between any two.
In this step 102, calculate document similarity between any two, at first need to obtain all speech that each document comprises, and then calculate document similarity between any two according to the speech that each document comprised.
At this,, obtain the speech that document comprises by document is made pauses in reading unpunctuated ancient writings and participle.Be specially, use simple punctuate method that document is made pauses in reading unpunctuated ancient writings, promptly scan entire document, if run into ".", "? ", "! " wait punctuation mark then to think the end of a sentence.After obtaining the sentence of document, each sentence is carried out participle.The method of participle can be utilized predefined dictionary, and each sentence is carried out reverse scanning, if a speech in the coupling dictionary, then output continues remaining part is carried out same reverse scanning, until participle finished in whole sentence then.
After obtaining the speech that each document comprises, can utilize the Jaccard formula to calculate any two document d iAnd d jBetween similarity.The Jaccard formula is as shown in the formula shown in (1).
J(d i，d j)＝|d i∩d j|/|d i∪d j| (1)
Wherein, | d i∩ d j| expression both had been included in document d iIn be included in document d again jIn the number of speech, i.e. document d iWith document d jThe number of the speech that comprises jointly; | d i∪ d j| expression only is included in document d iIn speech number, only be included in document d jIn the number and the document d of speech iWith document d jThe number sum of the speech that comprises jointly.| d i∪ d j| value can be by calculating document d iIn the number of the speech that comprised add document d jIn the number of the speech that comprised deduct document d again iWith document d jThe number of the speech that comprises jointly obtains.
For example, document d iBe " example document ", document d jBe " scanned document program ", suppose here the segmenting method of each Chinese character as a speech, then document d iComprise 4 speech, document d jComprise 6 speech.At this moment, | d i∩ d j|=2, | d i∪ d j|=4+6-2=8, therefore, document d iWith document d jBetween similarity be J (d i, d j)=| d i∩ d j|/| d i∪ d j|=2/8=0.25.
For another example, if document d iWith document d jWithout any the speech that comprises jointly, document d then iWith document d jBetween similarity be zero.If document d iWith document d jContent identical, document d then iWith document d jBetween similarity be 1.
Step 103 is obtained the correlativity evaluation for the document of Search Results.
The 103 described correlativity evaluations of this step are meant that the document to Search Results provides the evaluation with the search intention degree of correlation, for example " very relevant " or evaluations such as " wide of the marks ".This evaluation can have a plurality of ranks, for example, according to the height of degree of correlation can provide " excellent ", " very ", " in ", 5 other evaluations of level such as " reaching ", " poor ".
It is explicit correlativity evaluation that correlativity evaluation of the present invention is preferably, and that is to say that the user provides explicit evaluation according to the document content degree relevant with search intention.The mode that the user provides explicit correlativity evaluation is specifically as follows, the user selects the one feedback from a plurality of evaluation options that search system provides, perhaps feed back the correlativity evaluation, perhaps directly provide the mark of evaluation by the user by the mode of answering the problem that search system proposed.
In this step, after obtaining the user and estimating,, convert the correlativity evaluation to corresponding numerical value for the ease of quantitative Analysis for the correlativity that document fed back of Search Results, for example, can be to these evaluation result normalizings to the numerical value between [1 ,+1].Specifically for instance, for above-mentioned 5 other evaluations of level " excellent ", " very ", " in ", the numerical value of " reaching ", " poor " normalizing is respectively "+1 ", "+0.5 ", " 0 ", " 0.5 ", " 1 ".When converting the correlativity evaluation to corresponding numerical value, be not limited to for example can also convert the numerical value of evaluation result normalizing between [1 ,+1] between [5 ,+5] numerical value.
The user can feed back explicit evaluation to the one or more documents in the document of Search Results.Do not feed back the document of explicit evaluation for the user, search system can just can obtain degree of correlation between these documents and the search intention according to the similarity between the document that obtains in the step 102, promptly as the correlativity evaluation that in fact obtains the user.
Step 104, according to the document of Search Results between any two similarity and to the correlativity evaluation of the document of Search Results, calculate the degree of correlation score of each document of Search Results.
File correlation is meant the degree of correlation of document and user search intent.The degree of correlation score of preferable each document that passes through the cycle calculations Search Results in the present embodiment is stablized up to the degree of correlation score of all documents, perhaps reaches the cycle index of regulation.After cycle calculations finished, the degree of correlation score of resulting document represented that the document is under the correlativity that the user fed back is estimated, with the degree of correlation of the true search intention of user.
Provide the detailed step that calculates the file correlation score below.At first define some parameters.To document d iThe correlativity evaluation value rf corresponding that is fed back with it iExpression, wherein, correlativity evaluation of estimate rf iFor example be that the value that normalizing obtains, i.e. rf are carried out in the correlativity evaluation i∈ [1 ,+1].As previously mentioned, the user not necessarily to all document feedback correlativity evaluations of Search Results, perhaps only to the evaluation of a document feedback correlativity, perhaps feeds back the correlativity evaluation to a plurality of documents.Have again, to document d iDegree of correlation score w iExpression.In the present embodiment, degree of correlation score is to upgrade by circulation to obtain, and in order to distinguish the degree of correlation score in the different cyclic processes, uses w i TRepresent the document d in the circulation the T time iDegree of correlation score, T=0 represents document d iThe initial value of degree of correlation score.
Figure 2 shows that the process flow diagram that calculates the file correlation score, comprise following step:
Step 201, the degree of correlation score of each document of initialization, shown in (2):
D ′ = ( ( d 1 , w 1 0 ) , . . . , ( d i , w i 0 ) , . . . , ( d N , w N o ) ) - - - ( 2 )
When initialization, if user feedback document d iThe correlativity evaluation, the document d then iThe initial value of degree of correlation score be correlativity evaluation of estimate rf i, promptly w i 0 = rf i ; If the user is to document d iDo not feed back the correlativity evaluation, then the document d iThe initial value of degree of correlation score can be made as default value, this default value for example is zero, promptly w i 0 = 0 .
In this step, it is I that maximum cycle index is set, and loop variable T is set to 1.
Step 202 is for each document d i, according to upgrading document d as shown in the formula (3) iDegree of correlation score:
w i T = Σ j = 1 N ( w j T - 1 * J ( d j , d j ) ) i∈[1，N] (3)
Wherein, w j T-1Be the document d that in last once circulation, obtains jDegree of correlation score, J (d j, d i) be document d jWith document d iBetween similarity, when i=j, owing to be identical document, so J (d j, d i)=1.
As the formula (3), when calculating the degree of correlation score of document, utilized the similarity between the document, like this, when the correlativity of a small amount of document of user feedback is estimated, obtain the close document of the document of correlativity evaluation to these and also can obtain similar correlativity evaluation, thus can be according to user's search intention, the document of Search Results is sorted and exports to the user.
Through type (3) obtains the degree of correlation score of all documents in this cyclic process, shown in (4):
D T = ( ( d 1 , w 1 T ) , . . . , ( d i , w i T ) , . . . , ( d N , w N T ) ) - - - ( 4 )
Step 203 from the degree of correlation score of upgrade all documents that obtain, is searched the highest score w Max, and to each document d iDegree of correlation score carry out normalizing and handle, promptly as shown in the formula shown in (5):
w i T = w 1 1 / w Max i∈[1，N] (5)
Step 204, judge whether to arrive maximum cycle index I, judge promptly whether loop variable T equals maximum cycle index I, and perhaps whether degree of correlation score is stable, if arrive maximum cycle index, perhaps degree of correlation score is stable, then enters step 205, otherwise, if the cycle index of no show maximum and degree of correlation score are unstable yet, then loop variable T is added 1, and return step 202, enter circulation next time.
In this step 204, whether the degree of correlation score of judging document stabilizes to, whether square sum of difference of judging the degree of correlation score that obtains in the degree of correlation score that obtains and last circulation in this circulation is less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise be judged to be stable.Perhaps, judge the degree of correlation score that obtains in the degree of correlation score that obtains and last circulation in this circulation difference square root sum square whether less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise be judged to be stable.More than two kinds of situations specifically be formulated as shown in the formula shown in (6) or (7):
&Sigma; i = 1 N ( w i T - w i T - 1 ) 2 < &epsiv; - - - ( 6 )
&Sigma; i = 1 N ( w i T - w i T - 1 ) 2 < &epsiv; - - - ( 7 )
Wherein, ε is the threshold value that is used to judge that degree of correlation score is whether stable.
In the present embodiment, will withdraw from the round-robin condition and be made as two, the one, judge whether cycle index arrives maximum cycle index, the 2nd, judge whether degree of correlation score is stable, and when satisfied wherein arbitrary condition, withdraw from circulation.Because the stable of degree of correlation score may be a very long process, therefore by suitably being set, maximum cycle I and threshold epsilon just can obtain metastable degree of correlation score, like this, according to the degree of correlation score that said method obtains, can determine the degree of correlation of each document and the true search intention of user preferably.
Step 105 is according to the degree of correlation score of each document of Search Results, to the output of sorting of the document of Search Results.
Owing to obtained preferably to reflect the degree of correlation score of the document of the true search intention of user by step 104, therefore, determine the priority of document ordering order according to the height of degree of correlation score, and export to the user after according to the height order of priority the document of Search Results being sorted.At this, the document that degree of correlation score is high more illustrates with user's search intention and gets over coupling that Sort Priority is then high more.
The Search Results optimizing process that provides by embodiment one, the user is really interested, the useful document that meets user search intent of user is returned to the user with high priority, thereby significantly reduce user's search time, make the user can be fast, obtain to meet the document of self search intention accurately, easily, improved search efficiency.And, in the present embodiment, according to the similarity between the search result document content calculating document, and utilize the degree of correlation of each document of the correlativity evaluation calculation Search Results of similarity between the document and user feedback, thereby avoid the user to participate in search procedure too much, and under the situation that a spot of correlativity of user feedback is estimated, search system also can obtain by the degree of correlation of calculating each document automatically and user search intent between degree of correlation, return the document that meets search intention for the user.
Embodiment two:
Fig. 3 is the synoptic diagram of among the embodiment two Search Results being optimized.Compare with embodiment one, in embodiment two, set up the similarity relation figure of document based on document content.Below in conjunction with Fig. 3, the process of Search Results optimization is described, but, does not describe in detail for the step identical with embodiment one.
Step 301, the key word of importing according to the user obtains Search Results.
Step 302 for the document of Search Results, is calculated document similarity between any two.
Step 303 for the document of Search Results, is set up the similarity relation figure of document based on the similarity between the document.
The described similarity relation figure of this step is with the document of the Search Results node as similarity relation figure, with document similarity between any two as the weight between the respective nodes of similarity relation figure.At this, similarity relation figure can adopt as shown in the formula (8) and represent.
G＝(D，V) (8)
Wherein, D=(d 1, d 2..., d i..., d j..., d N), being meant that each document of Search Results is distinguished corresponding node, its number is N with the number of documents of Search Results.
V represents the weight matrix between the node, can be expressed as:
Wherein, V " i, j "=J (d i, d j), two node d among the expression similarity relation figure iAnd d jBetween weight, its value equals two node d iAnd d jDistinguish similarity between the corresponding document.Therefore, weight matrix V can be called the similarity relation matrix between the document again.
Step 304 is obtained the correlativity evaluation for the document that searches out.
Step 305 according to the similarity relation figure of the document of Search Results with to the correlativity evaluation of the document of Search Results, is calculated the degree of correlation score of each document of Search Results.
In the present embodiment, the preferable degree of correlation score of passing through each node among the cycle calculations similarity relation figure is stablized up to the degree of correlation score of all nodes, perhaps reaches the cycle index of regulation.Concrete steps are with step shown in Figure 2, but distinctive points only is that upgrade on the formula of degree of correlation score, promptly the similarity relation matrix among the similarity relation figure that obtains with step 303 is specially in present embodiment two, w i T = &Sigma; j = 1 N ( w j T - 1 * V [ j , i ] ) i∈[1，N] (10)
Wherein, V[j, i] be node d jWith node d iBetween weight, its value equals the document d corresponding with described node jWith document d iBetween similarity J (d j, d i), when i=j, owing to be identical node, so V[j, i]=1.
Step 306 is according to the degree of correlation score of each document of Search Results, to the output of sorting of the document of Search Results.
By embodiment two, reaching outside the embodiment one described technique effect, also, be convenient to by the above-mentioned search optimizing process of computer realization owing to adopt the similarity relation figure of document.
Above-described search optimization method not only can be applied to internet hunt, WDS, company information search, can also be applied to the search application system on instant messaging (IM) equipment, mobile phone mobile device and the handheld device.
The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (6)

1. a method that is used to search for is characterized in that, this method comprises:
Obtain correlativity evaluation for the document of Search Results;
According to the correlativity evaluation of the document of Search Results and the document content of Search Results, the document of Search Results is sorted, specifically comprise:
According to the document content of Search Results, calculate the document similarity between any two of Search Results;
Obtain the similarity relation figure of the document of Search Results, the node of described similarity relation figure is the document of Search Results, and the weight between the node of described similarity relation figure is the similarity between the corresponding document of described node;
According to the correlativity evaluation of the document of Search Results and the document similarity relation figure of Search Results, the degree of correlation of each document of initialization, for obtaining the document that correlativity is estimated, the initial value of its degree of correlation is to estimate corresponding value with correlativity, for the document that does not have correlativity to estimate, the initial value of its degree of correlation is a default value;
The document that utilizes Search Results between any two similarity and the initial value of the degree of correlation or last cyclic process in the degree of correlation that obtains, upgrade the degree of correlation that obtains each document in this cyclic process;
Judge whether to reach the cycle index of regulation, whether the degree of correlation of judging each document is stable, if reaching the cycle index of regulation or the degree of correlation of each document stablizes, then with the degree of correlation of each document of obtaining in this circulation the degree of correlation, otherwise return the step of carrying out the degree of correlation of upgrading each document as each document;
Sort according to the degree of correlation of resulting each document document to Search Results.
2. method according to claim 1 is characterized in that, the step of described calculating document similarity between any two comprises:
Obtain the speech that each document comprises, obtain the number of the speech that each document comprises;
When first document in calculating search result document and the similarity between second document,
Obtain the number of the speech that described first document and second document comprise jointly;
The number of the speech that is comprised with described first document and the number sum of the speech that second document is comprised deduct the value that the number of the speech that described first document and second document comprise jointly obtains again, remove the resulting value of number of the speech that described first document and second document comprise jointly, as the similarity between described first document and second document.
3. method according to claim 1 is characterized in that, describedly judges whether the degree of correlation of each document stabilizes to:
Whether square sum of difference of judging the degree of correlation that obtains in the degree of correlation that obtains and last circulation in this circulation less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise is judged to be stable;
Perhaps, judge the degree of correlation that obtains in the degree of correlation that obtains and last circulation in this circulation difference square root sum square whether less than the threshold value of regulation, if less than the threshold value of regulation then be judged to be stablely, otherwise be judged to be stable.
4. method according to claim 1 is characterized in that,
When the circulation time first time, the degree of correlation of each document is in described this cyclic process,
Wherein, w j 0Document d jThe initial value of the degree of correlation, J (d j, d i) be document d jWith document d iBetween similarity, w i TBe the T time document d in the circulation iDegree of correlation score, N is the number of documents of Search Results, T be the T time the circulation;
When the above circulation time second time, the degree of correlation of each document is in described this cyclic process,
Figure F2007101882729C00022
Wherein, w j T-1Be the document d that in last once circulation, obtains iThe degree of correlation, w i TBe the T time document d in the circulation iDegree of correlation score, J (d j, d i) be document d jWith document d iBetween similarity, N is the number of documents of Search Results, T be the T time the circulation.
5. method according to claim 1, it is characterized in that, described renewal obtains after the degree of correlation of each document in this cyclic process, before carrying out described judgement, further comprise: search maximal value in the degree of correlation of each document that from this circulation, obtains, and according to this maximal value the degree of correlation of each document is carried out normalizing and handle.
6. method according to claim 1, it is characterized in that, the correlativity evaluation of described document for Search Results is selected one of them option feedback by the user from a plurality of evaluation options that search system provides, perhaps feed back by the mode of answering the problem that search system proposed, perhaps the mode that directly provides evaluation score by the user is fed back.
CN2007101882729A 2007-11-30 2007-11-30 Searching method Expired - Fee Related CN101169797B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2007101882729A CN101169797B (en) 2007-11-30 2007-11-30 Searching method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2007101882729A CN101169797B (en) 2007-11-30 2007-11-30 Searching method

Publications (2)

Publication Number Publication Date
CN101169797A CN101169797A (en) 2008-04-30
CN101169797B true CN101169797B (en) 2010-04-07

Family

ID=39390422

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2007101882729A Expired - Fee Related CN101169797B (en) 2007-11-30 2007-11-30 Searching method

Country Status (1)

Country Link
CN (1) CN101169797B (en)

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102236677A (en) * 2010-04-28 2011-11-09 北京大学深圳研究生院 Question answering system-based information matching method and system
CN101968799B (en) * 2010-09-21 2012-02-08 百度在线网络技术（北京）有限公司 Search engine-based user interaction method and system
US9251185B2 (en) 2010-12-15 2016-02-02 Girish Kumar Classifying results of search queries
WO2012097504A1 (en) * 2011-01-18 2012-07-26 Google Inc. Automated answers to online questions
CN102682040A (en) * 2011-03-16 2012-09-19 日电（中国）有限公司 Device and method for calculating importance of documents
CN103838773A (en) * 2012-11-26 2014-06-04 百度在线网络技术（北京）有限公司 User judgment method and device for search result
CN103034709B (en) * 2012-12-07 2017-05-31 北京海量融通软件技术有限公司 Retrieving result reordering system and method
CN104281619A (en) * 2013-07-11 2015-01-14 鸿富锦精密工业（深圳）有限公司 System and method for ordering search results
CN103646092B (en) * 2013-12-18 2017-07-04 孙燕群 Based on the method for sequencing search engines that user participates in
CN104572998B (en) * 2015-01-07 2017-09-01 北京云知声信息技术有限公司 Question and answer order models update method and device for automatically request-answering system
CN104933116A (en) * 2015-06-04 2015-09-23 科大讯飞股份有限公司 Grading method and device of movie and television relevant information
CN105159932B (en) * 2015-08-07 2019-06-21 中车青岛四方机车车辆股份有限公司 A kind of data retrieval engine and ordering system and method
CN106649407A (en) * 2015-11-04 2017-05-10 阿里巴巴集团控股有限公司 Retrieval result obtaining method and apparatus
CN105574095A (en) * 2015-12-10 2016-05-11 成都陌云科技有限公司 Context-sensitive intelligent estimation method for user behaviors
JP2020030634A (en) * 2018-08-23 2020-02-27 株式会社Ｓｃｒｅｅｎホールディングス Search device, search method, and search program
CN111125322B (en) * 2019-11-19 2021-02-12 北京金堤科技有限公司 Information searching method and device, electronic equipment and storage medium
CN114185500B (en) * 2021-12-14 2024-04-02 深圳市润天智数字设备股份有限公司 Control method and device for printing job

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1465197A (en) * 2001-04-25 2003-12-31 皇家菲利浦电子有限公司 Apparatus and method for combining random set of video features in a non-linear scheme to best describe perceptual quality of video sequences using heuristic search methodology
CN1818908A (en) * 2006-03-16 2006-08-16 董崇军 Feedbakc information use of searcher in search engine
CN101046804A (en) * 2006-03-30 2007-10-03 国际商业机器公司 Method for searching order in file system and correlation search engine

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1465197A (en) * 2001-04-25 2003-12-31 皇家菲利浦电子有限公司 Apparatus and method for combining random set of video features in a non-linear scheme to best describe perceptual quality of video sequences using heuristic search methodology
CN1818908A (en) * 2006-03-16 2006-08-16 董崇军 Feedbakc information use of searcher in search engine
CN101046804A (en) * 2006-03-30 2007-10-03 国际商业机器公司 Method for searching order in file system and correlation search engine

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Tingshao Zhu et al..Off-line evaluation of recommendation functions.User Modeling 2005. 10th International Conference.2005,337-41. *

Also Published As

Publication number Publication date
CN101169797A (en) 2008-04-30

Similar Documents

Publication Publication Date Title
CN101169797B (en) Searching method
CN106202382B (en) Link instance method and system
CN103186574B (en) A kind of generation method and apparatus of Search Results
CN105955976B (en) A kind of automatic answering system and method
US20170213157A1 (en) Method and system to provide related data
CN1977261B (en) Method and system for word sequence processing
RU2671047C2 (en) Search tables understanding
CN103218436B (en) A kind of Similar Problems search method and device that merges class of subscriber label
CN100545847C (en) A kind of method and system that blog articles is sorted
CN1609845A (en) Method and apparatus for improving readability of automatic generated abstract by machine
JP2011227889A (en) Method for calculating semantic similarity between message and conversation based on extended entity extraction
CN107145545A (en) Top k zone users text data recommends method in a kind of location-based social networks
CN101625680A (en) Document retrieval method in patent field
CN103838735A (en) Data retrieval method for improving retrieval efficiency and quality
US9971828B2 (en) Document tagging and retrieval using per-subject dictionaries including subject-determining-power scores for entries
CN101546331A (en) System and method for acquiring characteristics favorable for retrieval and evaluating value of related things
CN112149422B (en) Dynamic enterprise news monitoring method based on natural language
Cao et al. Mashup service recommendation based on usage history and service network
CN102890690B (en) Target information search method and device
CN110377684A (en) A kind of spatial key personalization semantic query method based on user feedback
CN108416019A (en) Conjunctive word method of adjustment and adjustment system
CN100481079C (en) Image extraction feedback method in web search
CN115391500A (en) Dialogue type information retrieval method based on pre-training language model
CN114912431A (en) Document searching method and device
CN100593783C (en) Method, system and device for acquiring appraisement of vocabulary semanteme

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
C17 Cessation of patent right
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20100407

Termination date: 20101130