CN108509610A - A kind of search method and system of activity and companion - Google Patents
A kind of search method and system of activity and companion Download PDFInfo
- Publication number
- CN108509610A CN108509610A CN201810294059.4A CN201810294059A CN108509610A CN 108509610 A CN108509610 A CN 108509610A CN 201810294059 A CN201810294059 A CN 201810294059A CN 108509610 A CN108509610 A CN 108509610A
- Authority
- CN
- China
- Prior art keywords
- activity
- companion
- similarity
- user
- movable
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention is suitable for Internet technology neighborhood, provide a kind of search method of activity and companion, the embodiment of the present invention is searched according to the inquiry user of input and searching keyword, obtain include it is described inquiry user pairing person and the searching keyword match activities as a result, the embodiment of the present invention merged activity with Peer referral and keyword query the advantages of.Its retrieval activities and companion couple consider related text relevant and the correlation of going together for inquiring user to companion's expected degree of going together with text file.The embodiment of the present invention is all considered the relationship of text similarity and activity and people when retrieval activities, can be inquired according to inquiry user and searching keyword, and the run time of single inquiry is second grade, and inquiry velocity is fast.
Description
Technology neighborhood
The invention belongs to Internet technology neighborhood more particularly to a kind of search methods and system of activity and companion.
Background technology
Movable social networks (EBSN) can use two-dimensional plot (G=<U,E,R>) simulate and, as described in Fig. 1, wherein U
User's set is represented, E deputy activity set, R represents the situation of user's participation activity.For example, r ∈ R=(u, e), u ∈ U, e ∈
E.Each of E activity e are related to a text file e. ψ of feature with description activity description, in general, this document vector table
Show, and movable member indicates to participate in the movable user.
Social networks based on activity, which flourishes, causes the very big concern of research circle.Many research work surround
Activity recommendation, place recommendation, friend recommendation and activity schedule and tissue.The activity occurred recently and Peer referral think recommended
Activity can also be target user recommend companion.However, these recommended technologies cannot be satisfied the real-time search need of user, together
When, traditional information retrieval technique is independently processed from activity and user, does not account for being combined activity with user and retrieve.
Invention content
Technical problem to be solved by the present invention lies in provide a kind of search method and system of activity and companion, it is intended to solve
Certainly existing recommended technology cannot be satisfied the real-time search need of user, and be the problem of being independently processed from activity and user.
The invention is realized in this way a kind of search method of activity and companion, including：
Step A receives inquiry user and searching keyword, obtains described in the activity that the inquiry user participated in and deposit
Inquire the active set of user；
Step B obtains the crucial companion of the inquiry user；
Step C determines activity to be matched according to the searching keyword, and the inquiry is calculated using matching similarity function
User and the crucial companion participate in the movable similarity upper limit of going together to be matched；
Step D judges whether the activity to be matched meets beta pruning condition according to colleague's similarity upper limit, if full
Foot, then export several results pair, if not satisfied, then calculating the neighborhood of the inquiry user；
Step E judges whether the neighborhood is sky, if so, return to step C, if it is not, then judging that the inquiry user is
It is no to there is crucial companion；
Step F matches the query object and the activity to be matched if the inquiry user has crucial companion
For movable companion couple, the matching similarity of the movable companion couple is calculated, by the movable companion to preserving to Priority Queues, is returned
Return step C；
Step G calculates worst colleague's similarity and most preferably goes together similar if crucial companion is not present in the query object
Degree judges whether to need beta pruning, if desired, then return to step according to worst colleague's similarity and the best similarity of going together
Rapid C is calculated if not needing and be whether there is pairing person；
Step H, however, it is determined that there are pairing persons, then matching similarity are calculated according to the pairing person, according to the matching phase
Like several pairs of degree output as a result, the result includes pairing person and the match activities of the inquiry user.
The present invention also provides a kind of searching systems of activity and companion, including：
Information receiving unit obtains the work that the inquiry user participated in for receiving inquiry user and searching keyword
Move and be stored in the active set of the inquiry user；
Companion's acquiring unit, the crucial companion for obtaining the inquiry user；
Upper limit computing unit utilizes matching similarity function for determining activity to be matched according to the searching keyword
It calculates the inquiry user and the crucial companion participates in the movable similarity upper limit of going together to be matched；
Beta pruning arithmetic element, for judging whether the activity to be matched meets beta pruning according to colleague's similarity upper limit
Condition, if satisfied, several results pair are then exported, if not satisfied, then calculating the neighborhood of the inquiry user；
Neighborhood judging unit, for judging whether the neighborhood is empty, if so, the upper limit computing unit is activated, if
It is no, then judge that the inquiry user whether there is crucial companion；
Companion's judging unit by the query object and described waits for if there is key companion for the inquiry user
Match activities matching is movable companion couple, calculates the matching similarity of the movable companion couple, by the movable companion to preserving
To Priority Queues, the upper limit computing unit is activated；If crucial companion is not present in the query object, worst colleague's phase is calculated
Like degree and best similarity of going together, judge whether to need to cut according to worst colleague's similarity and the best similarity of going together
Branch, if desired, then activate the upper limit computing unit, if not needing, calculate and whether there is pairing person；
As a result output unit is used to if it is determined that there are pairing person, then calculate matching similarity according to the pairing person, according to
The matching similarity exports several pairs as a result, the result includes pairing person and the match activities of the inquiry user.
Compared with prior art, the present invention advantageous effect is：The embodiment of the present invention is according to the inquiry user of input and Cha
Ask keyword searched, obtain include it is described inquire user pairing person and the searching keyword match activities knot
The advantages of fruit, the activity of having merged of the embodiment of the present invention is with Peer referral and keyword query.Its retrieval activities and companion couple, it is comprehensive
Consider related text relevant and the correlation of going together for inquiring user to companion's expected degree of going together with text file.The present invention
Embodiment is all considered the relationship of text similarity and activity and people when retrieval activities, can be used according to inquiry
Family and searching keyword are inquired, and the run time of single inquiry is second grade, and inquiry velocity is fast.
Description of the drawings
Fig. 1 is the two-dimensional plot that the prior art provides；
Fig. 2 is activity provided in an embodiment of the present invention and user's table；
Fig. 3 is the flow chart of a kind of activity and the search method of companion provided in an embodiment of the present invention；
Fig. 4 is the flow chart provided in an embodiment of the present invention for calculating pairing person；
Fig. 5 a to Fig. 5 d are the comparison of test results figures that the search method provided through the embodiment of the present invention is tested；
Fig. 6 is the structural schematic diagram of a kind of activity and the searching system of companion provided in an embodiment of the present invention.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Colleague's similarity p (u*, e, u) u* pairs of expected degree for removing movable e jointly with user u of measure user.Such as following formula public affairs
Formula (1) is determined by two factors.First, user usually wants the movable similar activity participated in once gone, because of people
Point of interest usually have certain specific features.This is the common attribute of many commending systems, and extensive for a long time
It uses.Second, people are usual and have people's activity of common interest.In embodiments of the present invention, the common interest of two people
Degree can be expressed as the activity that they participated in jointly.
WhereinIfGoing together, similarity p (u*, e, u) is nonsensical, this expression does not have
Peer referral removes activity e to u*.This is the restrictive condition for defining colleague's similarity.The purpose of the embodiment of the present invention is to carry
The efficiency of high calculating activity and companion couple, and the purpose that colleague's similarity is not this work is improved, while nor this grinds
Study carefully future direction.There are three parameters in formula (1)：Target user u*, movable e and pairing user u, are meant that target user
U* may be with the pairing common activity e of user u.The range of p (u*, e, u) is [0,1].The bigger expression mesh of value of p (u*, e, u)
The hope for marking user u* and the common activity e of pairing user u is bigger.In formula (1), the neighbour of activity and user to (u*, e)
Domain N (u*, e) is by a series of activity { eiDefine, wherein each activity eiMeet two following conditions：(1) u* was participated in
ei.(2) activity e and eiSimilarity s (e, ei) it is not less than threshold tau.Given pairing user u, if u* and u participated in activity
eiSo b (u*, e, u)=1, otherwise b (u*, e, u)=0.Determine formula (1) is movable e with activity and user to (u*, e)
Neighborhood in each activity eiSimilarity sum.U in movable e and neighborhood N (u*, e) each of was participated in activity e by denominatori's
Similarity is added.Colleague's similarity is not symmetrical, such as p (u*, e, u) ≠ p (u, e, u*).
Colleague's similarity can be stated with two words, i.e.,：First, if u* participated in activity it is similar with activity e,
Movable e is the candidate active that user u* will be participated in, such asSecondly, if user u was participated in N (u*, e)
Certain activities, then user u is candidate companions of the user u* about movable e.
In embodiments of the present invention, the search method of really preceding k most correlated activations and companion, i.e. kEP, wherein look into
Ask Q=(k, uq,yq) include following three parameter：(1) the number k of the activity and companion couple of inquiry.(2) inquiry user uq。(3)
One group polling keyword yq.Enable t (yq, e, y) and text similarity as activity e about searching keyword.KEP query result packets
Include the result pair of score highest activity and companion that k calculates the fractional function f provided according to formula 2.The result centering
Activity is unique, but different activities may be matched with the same companion.Fractional function f consider simultaneously text similarity and
Colleague's similarity.If there is the identical situation of score, optional one.Inquire user uqIt is not necessarily intended to participate in the work in result
It is dynamic.In order to which without loss of generality, the weighted sum of text similarity and similarity of going together is utilized in fractional function f so that the present invention is real
The search method for applying example offer can be used for any monotonic function about text similarity and similarity of going together.KEP inquiries are tasted
Examination finds activity with companion to (e, u), wherein activity is related to searching keyword, inquiry user is likely to jointly join with companion u
Movable e, formula (2) such as following formula is added to indicate：
f(uq,yq, e, u) and=α t (yq,e)+(1-α)·p(uq,e,u)
s.t.t(yq,e)∈(0,1]∧p(uq,e,u)∈(0,1]； (2)
In EBSN, some activities may periodically be held.Such as weekly or one month primary.Certain user may
The same activity is participated in repeatedly, therefore, in kEP query results, it is understood that there may be the activity that some inquiry users had been participating in,
These activities are not excluded in inventive embodiments, perhaps user is interested in them to participate in again because inquiring.
Top-k activities and the search method (kEP) of companion are proposed in the embodiment of the present invention, the activity of having merged is pushed away with companion
The advantages of recommending model and keyword query.When being retrieved while considering movable text relevant and correlation of going together, it is preceding
Person comes from text file, and the latter indicates expected degree of the inquiry user to the companion that goes together.In order to more efficiently obtain kEP inquiries
As a result, the innovatory algorithm that the embodiment of the present invention proposes uses three Pruning strategies and crucial companion's lookup method.
Based on above statement, an embodiment of the present invention provides the search method of a kind of activity and companion as shown in Figure 3, packets
It includes：
S301 receives inquiry user and searching keyword, obtains described in the activity that the inquiry user participated in and deposit
Inquire the active set of user；
S302 obtains the crucial companion of the inquiry user；
S303 determines activity to be matched according to the searching keyword, and the inquiry is calculated using matching similarity function
User and the crucial companion participate in the movable similarity upper limit of going together to be matched；
S304 judges whether the activity to be matched meets beta pruning condition according to colleague's similarity upper limit, if satisfied,
Several results pair are then exported, if not satisfied, then calculating the neighborhood of the inquiry user；
S305 judges whether the neighborhood is empty, if so, S303 is returned to, if it is not, whether then judging the inquiry user
There are crucial companions；
S306 matches the query object and the activity to be matched if the inquiry user has crucial companion
For movable companion couple, the matching similarity of the movable companion couple is calculated, by the movable companion to preserving to Priority Queues, is returned
Return S303；
S307 calculates worst colleague's similarity and most preferably goes together similar if crucial companion is not present in the query object
Degree judges whether to need beta pruning, if desired, then return according to worst colleague's similarity and the best similarity of going together
S303 is calculated if not needing and be whether there is pairing person；
S308, however, it is determined that there are pairing persons, then calculate matching similarity according to the pairing person, similar according to the matching
Several pairs of degree output is as a result, the result includes pairing person and the match activities of the inquiry user.
In step S301, searching keyword corresponds to the activity stored in activity database, in specific retrieving
In, searching system searches the activity in certain similarity according to searching keyword input by user in activity database, and
Most like activity is determined in the activity for the certain similarity searched, which has with the activity that user participated in
The most like activity is preserved into active set, and preserves the movable similarity of the most like degree by highest similarity,
The most like activity is finally exported in display field.
The embodiment of the present invention proposes a kind of search method of the activity and companion of top-k, the activity of having merged and Peer referral
And the advantages of keyword query.Its retrieval activities and companion couple, consider text relevant related with text file and look into
Ask colleague correlation of the user to companion's expected degree of going together.For example, user Mary wants to look for and " music performance " related activity
With possible companion.Therefore search method provided in an embodiment of the present invention had both been different from only retrieving correlated activation without retrieving companion
Keyword query, also different from based on historical data, lack the activity of flexibility and Peer referral that user searches in real time.
The input of kEP inquiries is a series of searching keywords and inquiry user, its retrieval highest k of matching similarity is to activity and together
Companion, if there is the identical situation of score, optional one.Activity and the score fractional function f of companion couple are calculated, it includes 1. living
The dynamic text relevant t for searching keyword；2. inquiring the similarity p that goes together of user and the common activity of unknown companion.
Therefore meet high text similarity when matching similarity function f selections activity and companion simultaneously and height is gone together similarity.
Improvement of the embodiment of the present invention in implementation process include：1, activity recommendation model and keyword query have been merged.
2, innovatory algorithm is proposed, which includes 1. shifting to an earlier date end condition 2. to being unsatisfactory for the activity beta pruning of requirement 3. to participating in
Companion beta pruning 4. crucial companion's technology.3, a series of experiments be used in it is on true data set the result shows that, should
Innovatory algorithm is much better than rudimentary algorithm.
In the embodiment of the present invention, in movable social networks all activities pass through index | |dIt is stored in storage database
In, it includes following two major parts：1, the vocabulary of the not repeated entries of all movable description files.2, each entry
A series of records.It is that a series of (id, w) are right, and wherein id represents an activity e, the description file e. ψ of this expression activity e
Including entry t；W is weights of the entry t in describing file e. ψ.EBSN two-dimensional plots G is stored in main memory.
The embodiment of the present invention is in specific implementation process, the termination in advance retrieved using end condition in advance, to subtract
Few calculation amount and raising calculating speed, wherein：
End condition includes in advance：The embodiment of the present invention retrieves activity related with searching keyword first, presses and inquiry
What crucial Word similarity successively decreased is ranked sequentially activity.When the result comprising activity and companion is more than k to number, item is terminated in advance
Part is set up, and obtained k is to result to as final result, the similarity activity lower than current active is all not required to consider.
In embodiments of the present invention, the lemma and beta pruning principle used includes：
Lemma 1, given inquiry user uqWith searching keyword yqAnd the movable similarity upper limit of going together of companion's participation is：
fub(uq, e) and=α t (ψq,e)+(1-α)·1； (3)
In formula (3), t expressions activity and searching keyword ψqSimilarity, α indicate weight, specific obtaining step
Including：Index is established to the activity in activity database using Lucene；According to the searching keyword ψq, using Lucene from
The sequence successively decreased according to text similarity in index retrieval activities one by one, obtain k activities to be matched；According to the inquiry
User and the searching keyword, are calculated using matching similarity function, obtain the inquiry user and companion participates in inspection
Rope each of arrives movable colleague's similarity upper limit f to be matchedub。
Beta pruning principle 2, given inquiry user uqWith searching keyword ψq, it is assumed that the sequence that activity is successively decreased by text similarity t
Retrieval.Enable fkIndicate that the score of the current kth of result centering big activity and companion couple, e indicate next activity that will be retrieved.Such as
Fruit fub(e)≤fk, then activity e and all activities that do not retrieve be not present in result centering, therefore by beta pruning.
In this beta pruning principle, each activity to be matched in k activities to be matched is calculated using matching similarity function f
With movable activity to be determined with companion to score, the activity includes that activity to be matched is crucial for inquiry to score with companion
The text relevant of word, and inquire the similarity p that goes together of user and the common activity of unknown companion；With fkIndicate k-th of work
Dynamic and companion is to score, and e indicates next activity that will be retrieved, if fub(e) ＞ fk, then by fkCorresponding activity and companion couple
It is saved in result centering, obtains several results pair；If fub(e)≤fk, then activity e and all activities that do not retrieve be not present in tying
Fruit centering by movable e beta prunings, and calculates the neighborhood N (u of inquiry userq, e), the neighborhood indicates the inquiry user uqIt went
Active set Eq and current active e similarity s (e, ei) it is more than the movable set of preset t.
The embodiment of the present invention proposes that does not have to the specific beta pruning principle for calculating pairing person, it may determine that an activity is
The no result that can be present in is in the result set of composition.In an experiment, the calculation amount of this pruning method is significantly less than calculating activity
Pairing person.
It is based on such as giving a definition and lemma：Worst colleague's similarity pw(e) (3 are defined) and shows that activity belongs to the big results of preceding k
The necessary condition of collection.Best colleague's similarity pub(e) (define 5) have estimated activity and its pairing person it is highest go together it is similar
Degree.Lemma 4 and 6 indicates p respectivelyw(e) and pub(e) feature.Beta pruning principle 7 is based on lemma 9 and 10, they can match not calculating
By the activity for the condition that is unsatisfactory under the premise of to person, i.e., the activity of k result centering is cut off before can not possibly appearing in, therefore beta pruning
Principle 7 can reduce calculation amount.
Define 3, worst colleague's similarity pw(e)：
Given inquiry user uqWith searching keyword ψq, enable fkIndicate point of the current kth of result centering big activity and companion
Number, the movable worst colleague's similarities of e are：
If lemma 4, ∨ u ∈ (U { uq}))(p(uq,e,u)≤pw(e)), movable e is not present in result centering.
Define 5, best colleague's similarity pub(e)：
With N (uq,e) indicate inquiry user uqWith the neighborhood of movable e, m=max | Ec||∨u∈(U\{uq})(Ec=N (uq,
e)∩Eu), wherein Eu is the movable set that u was participated in, and best colleague's similarity of e is：
Wherein, | TopM (uq, e) |=m,Meanwhile ∨ ei∈TopM(uq,e)∨ej∈(N
(uq,e)\TopM(uq,e))(s(e,ei)≥s(e,ej))。
Lemma 6, e best colleague's similarity pub(e) be movable e and pairing person similarity of going together max-thresholds.
Beta pruning principle 7, given inquiry user uqWith searching keyword ψq.To movable e, if pw(e)≥pub(e), movable e
It is not present in result team, therefore by beta pruning.
According to the definition of colleague's similarity, all neighborhood N (u have been participated in if there is a user uq,e) in activity, together
Row similarity p (uq,e,up) be equal to 1, that is, p maximum value.It says more bluntly, user u is the pairing user of movable e.According to
This observation, the embodiment of the present invention are gathered by defining 8 to define the crucial companion of user u.If inquiring user u, there are one close
Key companion, then any candidate active can be matched with crucial companion, and similarity of going together is 1 (lemma 9).In other words,
It can more directly to activity and companion to there is the inquiry user of crucial companion to calculate preceding k.For example, using next activity function is obtained
K activity before directly searching, then as a result with crucial companion pairing.In this way, calculation amount substantially reduces.
Define 8, key companion set KP (u):
With EuAnd Eu’Indicate the activity that user u and user u ' was respectively attended, ifThen user u ' is user u
Crucial companion.Key is defined as follows with coset
Lemma 9, given inquiry user uq, ∨ e ∈ E ∨ u ∈ KP (uq)∨u'∈U\(KP(uq)∪{uq})(p(uq,e,u)
=1 >=p (uq,e,u'))。
The crucial of user with coset may include multiple users, according to the definition that kEP is inquired, as a result in activity be not heavy
It is multiple, therefore the embodiment of the present invention need not be by each user that crucial companion concentrates and certain active pair.The crucial companion of selection
Colleague's similarity of any user of concentration is all 1.In an experiment, only one user of arbitrary selection need to be concentrated in crucial companion.
The efficient retrieval of user：
Given inquiry user uqWith movable e, the colleague highest pairing person of similarity is found out, then result is formed with movable e
To (e, u).With including data structure as the membership table of activity-user in the present embodiment, to each activity, its membership table
It is made of (u, num), user sorts by the sequence that num successively decreases, and num indicates the activity number that user u was participated in.Membership table is as schemed
Shown in 2, movable e4There are two user u4And u5, user u4Participated in 5 activities, u5Participated in 3 activities.Calculating shown in Fig. 4
Pairing person gives the flow of user's efficient retrieval.Inquire user uq, activity e and neighborhood N (uq, e) and it is used as parameter, it returns and meets
That highest pairing user of similarity of going together.To neighborhood N (uq, e) in movable membership table searched, meet u ≠ uqBefore
Put, in membership table obtain first couple of num it is maximum (u, num), if two (u, num) to num be all the largest, that
First take similarity s (e, ei) big movable membership table (u, num).Then colleague similarity p (u are calculatedq, e, u), it will work as
Before (u, num) removed from membership table.If p (uq, e, u) and it is more than colleague's similarity p1, then p is updated1, and active user u is the most
Candidate pairing person.Colleague's similarity max-thresholds p is calculated againr(lemma 10), it indicates the most Datong District for being left user in membership table
Row similarity.If current candidate pairing person upP1More than or equal to pr, then upIt is so that p (uq, e, u) and the maximum user of value (cuts
Branch principle 11), and return to it.Otherwise third step is returned to obtain lower a pair of (u, num).
Lemma 10：Given inquiry user uqWith movable e, current pairing person and quantity are to for (u, num), enabling x=min
{num,|N(uq,e)|},TopX(uq, e) and include neighborhood N (uq, e) in similarity s (e, ei) maximum preceding x activity { ei}.It is adjacent
Domain N (uq, e) in all movable membership tables colleague's similarity max-thresholds of user be：
Beta pruning principle 11 enables p1As current candidate companion upColleague's similarity, if p1≥pr, user upAs activity
The pairing person of e returns, then pairing person will not be become by being left user in membership table to be checked, therefore by beta pruning.
The embodiment of the present invention is further illustrated with reference to Fig. 2：
The result pair including activity with companion is stored in Priority Queues.The activity function participated in is obtained to be used according to activity
Family two-dimensional plot G returns to uqThe movable set participated in.It obtains crucial companion's function and returns to inquiry user uqCrucial companion.It obtains
Take next activity function from index | |dIn the sequence successively decreased by text similarity retrieval activities and return one by one.To each inspection
The movable e that rope arrives calculates score upper bound fub(lemma 1).Judge whether beta pruning with beta pruning principle 2 again, if meeting beta pruning, i.e.,
The big result of kth is to fkMore than score upper bound fub, then k is to result before terminating and returning.Otherwise neighborhood N (u are calculatedq, e), it is to look into
Ask the active set E that user wentqNeutralize current active e similarity s (e, ei) it is more than or equal to the movable set of threshold tau.If adjacent
Domain N (uq, e) and it is sky, then colleague's similarity is nonsensical, therefore current active e will not be used as and inquire user and may participate in
Activity simultaneously abandons it, is circulated back to S303 and obtains next activity function.If neighborhood is not sky, using crucial companion's algorithm, if
Inquiry user possesses crucial companion up, then colleague similarity p (uq,e,up) value is 1, then calculates movable companion to (e, up)
Matching similarity f is further continued for retrieving and handling next activity by movable companion to Priority Queues is added.So only with inquiry k
A activity, by crucial companion upK is formed one by one to result pair with k activity.If inquiring the no crucial companion of user,
Worst colleague's similarity and best similarity of going together are calculated, then judges whether beta pruning with beta pruning principle 7, is returned to if beta pruning
S303.If the not beta pruning activity, algorithm as shown in Figure 4 calculates pairing person.Repeat the circulation portions in the algorithm that Fig. 4 is provided
Point, until meeting beta pruning principle 11.The highest k of matching similarity f are finally returned in Priority Queues to activity and companion's result pair.
Specifically, the step of calculating pairing person shown in Fig. 4 includes：
S401, to neighborhood N (uq, e) each of activity search membership table respectively；
S402 calculates current colleague's similarity max-thresholds p according to the membership table foundr,
Wherein (u, num) indicates candidate companion to be determined and quantity pair in the membership table, enable x=min num, | N (uq, e) | }, collection
Close TopX (uq, e) and include neighborhood N (uq, e) in similarity s (e, ei) maximum preceding x activity { ei}；
S403 enables p1As current candidate companion upColleague's similarity, if p1≥pr, then current candidate companion upAs work
The pairing person of dynamic e；
S404, if p1＜ pr, then obtain lower a pair of (u, num), and the colleague of the candidate companion in lower a pair of (u, num)
Similarity is more than p1When, update p1, and using candidate companion as candidate pairing person；
S405 removes current (u, num) from the membership table, and obtains lower a pair of (u, num), obtain num and |
N | smaller value, calculate colleague similarity degree max-thresholds pr, and return to step S403.
In the prior art, the development of commending system and information retrieval system is more independent, and the embodiment of the present invention is creative
Ground links up two aspects.The embodiment of the present invention is when retrieval activities by the pass of text similarity and activity and people
System all considers into.The embodiment of the present invention is all second grade to the run time that single is inquired, in one minute, when fast
Retrieval can be completed in one second.
The embodiment of the present invention is further illustrated below by specific experimental data：
Experimental data comes from www.meetup.com, including 224238 activities, 7822965 users, and experimental result is as schemed
Shown in 5a to Fig. 5 d.Basic skills is not with any Pruning strategy, improved method beta pruning 2 and 11.Complete improve has used institute
The efficient retrieval method of some beta prunings 2,7,11 and user, α indicate in fractional function f, the weight of text similarity, accordingly
The weight on ground, similarity of going together is exactly 1- α.τ indicates the threshold value of the similarity of the activity and current active that will be calculated,
It is more than namely this threshold value, can just considers this activity.Default parameters is：K is 10,3 searching keywords, α 0.5, phase
It is 0.3 like degree threshold tau.With control variate method, k 1,10,20,30；Searching keyword number takes 1,2,3,4,5；α takes 0.1,
0.3,0.5,0.7,0.9；τ takes 0.1,0.2,0.3,0.4,0.5.There are three figures for each variable:Average operating time, average computation
How many people, average computation how many activity, all results are all the average value of 100 query results.The cylindricality of white indicates
Not plus the methods of any beta pruning, grey expression have only used beta pruning principle 2 and 11, black to indicate to have used whole above methods.Knot
Fruit shows for either run time or calculation amount, promotion effect all highly significants of the embodiment of the present invention.
Fig. 6 shows the searching system of a kind of activity and companion provided in an embodiment of the present invention, including：
Information receiving unit 601 obtains what the inquiry user participated in for receiving inquiry user and searching keyword
Activity and the active set for being stored in the inquiry user；
Companion's acquiring unit 602, the crucial companion for obtaining the inquiry user；
Upper limit computing unit 603 utilizes matching similarity letter for determining activity to be matched according to the searching keyword
Number calculates the inquiry user and the crucial companion participates in the movable similarity upper limit of going together to be matched；
Beta pruning arithmetic element 604, for judging whether the activity to be matched meets according to colleague's similarity upper limit
Beta pruning condition, if satisfied, several results pair are then exported, if not satisfied, then calculating the neighborhood of the inquiry user；
Neighborhood judging unit 605, for judging whether the neighborhood is empty, if so, the upper limit computing unit is activated,
If it is not, then judging that the inquiry user whether there is crucial companion；
Companion's judging unit 606, if there is key companion for the inquiry user, by the query object and described
Activity matching to be matched is movable companion couple, calculates the matching similarity of the movable companion couple, by the movable companion to protecting
It deposits to Priority Queues, activation upper limit computing unit 603；If crucial companion is not present in the query object, worst colleague is calculated
Similarity and best similarity of going together judge whether needs according to worst colleague's similarity and the best similarity of going together
Beta pruning, if desired, then activate upper limit computing unit 603, if not needing, calculate and whether there is pairing person；
As a result output unit 607 are used to if it is determined that there are pairing person, then calculate matching similarity according to the pairing person,
Several pairs are exported as a result, the result includes pairing person and the match activities of the inquiry user according to the matching similarity.
Further, activity indicates that the inquiry user is with u with eqIt indicates, the searching keyword is with ψqIt indicates, with KP
(u) the inquiry user u is indicatedqThe same coset of key, with EuAnd Eu’Indicate the activity that user u and user u ' was respectively attended,
IfThen user u ' is the key that user u companions, i.e.,：
Wherein, U indicates the set of total user, uiIndicate i-th of user；
Upper limit computing unit 603 is specifically used for：
Index is established to the activity in activity database using Lucene, according to the searching keyword ψq, utilize
The sequence that Lucene successively decreases from the index according to text similarity retrieval activities one by one, obtain k activities to be matched, according to
The inquiry user and the searching keyword, are calculated using matching similarity function, obtain the inquiry user and same
With participating in each of retrieving movable colleague's similarity upper limit f to be matchedub, i.e.,：
fub(uq, e) and=α t (ψq, e) and+(1- α) 1,
Wherein t expressions activity and searching keyword ψqSimilarity, α indicate weight；
Beta pruning arithmetic element 604 is specifically used for：
First, each activity to be matched and work to be determined in k activities to be matched are calculated using matching similarity function f
With companion to score, the activity includes text phase of the activity to be matched for searching keyword to score with companion for dynamic activity
Guan Xing, and inquire the similarity p that goes together of user and the common activity of unknown companion；
Secondly, with fkK-th of activity and companion are indicated to score, e indicates next activity that will be retrieved, if fub(e)
＞ fk, then by fkCorresponding activity, to being saved in result centering, obtains several results pair with companion；
Finally, if fub(e)≤fk, then activity e and all activities that do not retrieve be not present in result centering, movable e cut
Branch, and calculate the neighborhood N (u of inquiry userq, e), the neighborhood indicates the inquiry user uqThe active set Eq that went and current
Movable e similarities s (e, ei) it is more than the movable set of preset t.
Further, if the inquiry user has crucial companion up, then similarity of going together p (uq,e,up) value is 1, then is counted
Calculation activity companion is to the matching similarity of (e, up), by the movable companion to preserving to Priority Queues；
With pw(e) worst colleague's similarity of expression activity e, fkIndicate the score of result centering kth big activity and companion,
Then：
With N (uq, e) and indicate inquiry user uqWith the neighborhood of movable e, m=max | Ec||∨u∈(U\{uq})(Ec=N
(uq,e)∩Eu), pub(e) best colleague's similarity of expression activity e, then：
Wherein, | TopM (uq, e) |=m,Meanwhile ∨ ei∈TopM(uq,e)∨ej∈(N
(uq,e)\TopM(uq,e))(s(e,ei)≥s(e,ej)), best colleague's similarity p of the activity eub(e) it is movable e and matches
To the max-thresholds of colleague's similarity of person；
Companion's judging unit 606 is specifically used for：
For movable e, if pw(e)≥pub(e), then activity e is not present in result centering, then by movable e beta prunings, and returns
Step C is returned, if pw(e) ＜ pub(e), then it calculates and whether there is pairing person；
Companion's judging unit is for executing following steps：
G1, to neighborhood N (uq,e) each of activity search membership table respectively；
G2 calculates current colleague's similarity max-thresholds p according to the membership table foundr,
Wherein (u, num) indicates candidate companion to be determined and quantity pair in the membership table, enable x=min num, | N (uq, e) | }, collection
Close TopX (uq, e) and include neighborhood N (uq, e) in similarity s (e, ei) maximum preceding x activity { ei}；
G3 enables p1As current candidate companion upColleague's similarity, if p1≥pr, then current candidate companion upAs activity
The pairing person of e；
G4, if p1＜ pr, then obtain lower a pair of (u, num), and colleague's phase of the candidate companion in lower a pair of (u, num)
It is more than p like degree1When, update p1, and using candidate companion as candidate pairing person；
G5 removes current (u, num) from the membership table, and obtains lower a pair of (u, num), obtain num and | N |
Smaller value, calculate colleague similarity degree max-thresholds pr, and return to step g3.
The embodiment of the present invention additionally provides a kind of terminal, including memory, processor and storage on a memory and are being located
The computer program run on reason device, which is characterized in that when processor executes computer program, realize one kind as shown in Figure 3
Activity and each step in the search method of companion.
A kind of readable storage medium storing program for executing is also provided in the embodiment of the present invention, is stored thereon with computer program, which is characterized in that
When the computer program is executed by processor, realize a kind of activity as shown in Figure 3 with it is each in the search method of companion
Step.
In addition, each function module in each embodiment of the present invention can be integrated in a processing module, it can also
That modules physically exist alone, can also two or more modules be integrated in a module.Above-mentioned integrated mould
The form that hardware had both may be used in block is realized, can also be realized in the form of software function module.
If the integrated module is realized in the form of software function module and sells or use as independent product
When, it can be stored in a computer read/write memory medium.Based on this understanding, technical scheme of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the present invention
Portion or part steps.And storage medium above-mentioned includes：USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can store journey
The medium of sequence code.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention made by within refreshing and principle.
Claims (10)
1. a kind of search method of activity and companion, which is characterized in that including：
Step A receives inquiry user and searching keyword, obtains the activity that the inquiry user participated in and is stored in the inquiry
The active set of user；
Step B obtains the crucial companion of the inquiry user；
Step C determines activity to be matched according to the searching keyword, and the inquiry user is calculated using matching similarity function
The movable similarity upper limit of going together to be matched is participated in the crucial companion；
Step D judges whether the activity to be matched meets beta pruning condition, if satisfied, then according to colleague's similarity upper limit
Several results pair are exported, if not satisfied, then calculating the neighborhood of the inquiry user；
Step E judges whether the neighborhood is sky, if so, return to step C, if it is not, then judging whether the inquiry user deposits
In crucial companion；
The query object and the activity matching to be matched are to live if the inquiry user has crucial companion by step F
Dynamic companion couple calculates the matching similarity of the movable companion couple, by the movable companion to preserving to Priority Queues, returns to step
Rapid C；
Step G calculates worst colleague's similarity and best similarity of going together if crucial companion is not present in the query object,
Judge whether to need beta pruning according to worst colleague's similarity and the best similarity of going together, if desired, then return to step
C is calculated if not needing and be whether there is pairing person；
Step H, however, it is determined that there are pairing persons, then matching similarity are calculated according to the pairing person, according to the matching similarity
Several pairs are exported as a result, the result includes pairing person and the match activities of the inquiry user.
2. search method as described in claim 1, which is characterized in that in stepb, activity is indicated with e, the inquiry user
With uqIt indicates, the searching keyword is with ψqIt indicates, the inquiry user u is indicated with KP (u)qThe same coset of key, with EuAnd Eu’
Indicate the activity that user u and user u ' was respectively attended, ifThen user u ' is the key that user u companions, i.e.,：
Wherein, U indicates the set of total user, uiIndicate i-th of user.
3. search method as claimed in claim 2, which is characterized in that the step C is specifically included：
Step C1 establishes index using Lucene to the activity in activity database；
Step C2, according to the searching keyword ψq, the sequence successively decreased according to text similarity from the index using Lucene
Retrieval activities one by one obtain k activities to be matched；
Step C3 is calculated according to the inquiry user and the searching keyword using matching similarity function, and institute is obtained
It states inquiry user and companion participates in each of retrieving the movable similarity upper limit f that goes together to be matchedub, i.e.,：
fub(uq, e) and=α t (ψq, e) and+(1- α) 1,
Wherein t expressions activity and searching keyword ψqSimilarity, α indicate weight.
4. search method as claimed in claim 3, which is characterized in that the step D is specifically included：
Step D1 calculates each activity to be matched and activity to be determined in k activities to be matched using matching similarity function f
Activity with companion to score, the activity includes that activity to be matched is related for the text of searching keyword to score to companion
Property, and inquire the similarity p that goes together of user and the common activity of unknown companion；
Step D2, with fkK-th of activity and companion are indicated to score, e indicates next activity that will be retrieved, if fub(e) ＞
fk, then by fkCorresponding activity, to being saved in result centering, obtains several results pair with companion；
Step D3, if fub(e)≤fk, then activity e and all activities that do not retrieve be not present in result centering, by movable e beta prunings,
And calculate the neighborhood N (u of inquiry userq, e), the neighborhood indicates the inquiry user uqThe active set Eq and current active gone
E similarity s (e, ei) it is more than the movable set of preset τ.
5. search method as claimed in claim 2, which is characterized in that in the step F, if the inquiry user has key
Companion up, then similarity of going together p (uq,e,up) value is 1, then calculates matching similarity of the movable companion to (e, up), by the work
Dynamic companion is to preserving to Priority Queues.
6. search method as claimed in claim 2, which is characterized in that in step G, with pw(e) the worst colleague of expression activity e
Similarity, fkIndicate the score of result centering kth big activity and companion, then：
With N (uq,e) indicate inquiry user uqWith the neighborhood of movable e, m=max | Ec||∨u∈(U\{uq})(Ec=N (uq,e)∩
Eu), pub(e) best colleague's similarity of expression activity e, then：
Wherein, | TopM (uq, e) |=m,Meanwhile ∨ ei∈TopM(uq,e)∨ej∈(N(uq,
e)\TopM(uq,e))(s(e,ei)≥s(e,ej)), best colleague's similarity p of the activity eub(e) it is movable e and pairing person
Colleague's similarity max-thresholds；
It is described according to worst colleague's similarity and the best similarity of going together judges whether to need the beta pruning include：
For movable e, if pw(e)≥pub(e), then activity e is not present in result centering, then by movable e beta prunings, and returns to step
Rapid C, if pw(e) ＜ pub(e), then it calculates and whether there is pairing person.
7. search method as claimed in claim 6, which is characterized in that calculating the step of whether there is pairing person includes：
Step G1, to neighborhood N (uq, e) each of activity search membership table respectively；
Step G2 calculates current colleague's similarity max-thresholds p according to the membership table foundr,
Wherein (u, num) indicates candidate companion to be determined and quantity pair in the membership table, enable x=min num, | N (uq, e) | }, collection
Close TopX (uq, e) and include neighborhood N (uq, e) in similarity s (e, ei) maximum preceding x activity { ei}；
Step G3, enables p1As current candidate companion upColleague's similarity, if p1≥pr, then current candidate companion upAs activity
The pairing person of e；
Step G4, if p1＜ pr, then obtain lower a pair of (u, num), and colleague's phase of the candidate companion in lower a pair of (u, num)
It is more than p like degree1When, update p1, and using candidate companion as candidate pairing person；
Step G5 removes current (u, num) from the membership table, and obtains lower a pair of (u, num), obtain num and | N |
Smaller value, calculate colleague similarity degree max-thresholds pr, and return to step G3.
8. a kind of searching system of activity and companion, which is characterized in that including：
Information receiving unit, for receiving inquiry user and searching keyword, the activity that the acquisition inquiry user participated in is simultaneously
It is stored in the active set of the inquiry user；
Companion's acquiring unit, the crucial companion for obtaining the inquiry user；
Upper limit computing unit is calculated for determining activity to be matched according to the searching keyword using matching similarity function
The inquiry user and the crucial companion participate in the movable similarity upper limit of going together to be matched；
Beta pruning arithmetic element, for judging whether the activity to be matched meets beta pruning item according to colleague's similarity upper limit
Part, if satisfied, several results pair are then exported, if not satisfied, then calculating the neighborhood of the inquiry user；
Neighborhood judging unit, for judging whether the neighborhood is empty, if so, the upper limit computing unit is activated, if it is not, then
Judge that the inquiry user whether there is crucial companion；
Companion's judging unit, if there is key companion for the inquiry user, by the query object and described to be matched
Activity matching is movable companion couple, calculates the matching similarity of the movable companion couple, by the movable companion to preserving to excellent
First queue activates the upper limit computing unit；If crucial companion is not present in the query object, worst colleague's similarity is calculated
With best similarity of going together, judge whether to need beta pruning according to worst colleague's similarity and the best similarity of going together,
If desired, then the upper limit computing unit is activated, if not needing, calculates and whether there is pairing person；
As a result output unit is used to if it is determined that there are pairing person, then matching similarity is calculated according to the pairing person, according to described
Matching similarity exports several pairs as a result, the result includes pairing person and the match activities of the inquiry user.
9. searching system as claimed in claim 8, which is characterized in that activity indicates that the inquiry user is with u with eqIt indicates, institute
Searching keyword is stated with yqIt indicates, the inquiry user u is indicated with KP (u)qThe same coset of key, with EuAnd Eu’Indicate user u and
The activity that user u ' was respectively attended, ifThen user u ' is the key that user u companions, i.e.,：
Wherein, U indicates the set of total user, uiIndicate i-th of user；
The upper limit computing unit is specifically used for：
Index is established to the activity in activity database using Lucene, according to the searching keyword ψq, using Lucene from institute
The sequence successively decreased according to text similarity in index retrieval activities one by one are stated, k activities to be matched are obtained, according to inquiry use
Family and the searching keyword, are calculated using matching similarity function, obtain the inquiry user and companion participates in retrieval
Each of arrive movable colleague's similarity upper limit f to be matchedub, i.e.,：
fub(uq, e) and=α t (ψq, e) and+(1- α) 1, wherein t expressions activity and searching keyword ψqSimilarity, α indicate power
Weight；
The beta pruning arithmetic element is specifically used for：
First, using matching similarity function f calculate in k activities to be matched each activity to be matched with it is to be determined movable
With companion to score, the activity includes that activity to be matched is related for the text of searching keyword to score to companion for activity
Property, and inquire the similarity p that goes together of user and the common activity of unknown companion；
Secondly, with fkK-th of activity and companion are indicated to score, e indicates next activity that will be retrieved, if fub(e) ＞ fk,
Then by fkCorresponding activity, to being saved in result centering, obtains several results pair with companion；
Finally, if fub(e)≤fk, then activity e and all activities that do not retrieve is not present in result centering, by movable e beta prunings, and
Calculate the neighborhood N (u of inquiry userq, e), the neighborhood indicates the inquiry user uqThe active set Eq and current active e gone
Similarity s (e, ei) it is more than the movable set of preset τ.
10. searching system as claimed in claim 9, which is characterized in that if the inquiry user has crucial companion up, then together
Row similarity p (uq,e,up) value is 1, then calculates matching similarity of the movable companion to (e, up), by the movable companion to protecting
It deposits to Priority Queues；
With pw(e) worst colleague's similarity of expression activity e, fkIndicate the score of result centering kth big activity and companion, then：
With N (uq,e) indicate inquiry user uqWith the neighborhood of movable e, m=max | Ec||∨u∈(U\{uq})(Ec=N (uq,e)∩
Eu), pub(e) best colleague's similarity of expression activity e, then：
Wherein, | TopM (uq, e) |=m,Meanwhile ∨ ei∈TopM(uq,e)∨ej∈(N(uq,
e)\TopM(uq,e))(s(e,ei)≥s(e,ej)), best colleague's similarity p of the activity eub(e) it is movable e and pairing person
Colleague's similarity max-thresholds；
Companion's judging unit is specifically used for：
For movable e, if pw(e)≥pub(e), then activity e is not present in result centering, then by movable e beta prunings, and returns to step
Rapid C, if pw(e) ＜ pub(e), then it calculates and whether there is pairing person；
Companion's judging unit is for executing following steps：
G1, to neighborhood N (uq,e) each of activity search membership table respectively；
G2 calculates current colleague's similarity max-thresholds p according to the membership table foundr,Its
In (u, num) indicate candidate companion to be determined and quantity pair in the membership table, enable x=min num, | N (uq, e) | }, set
TopX(uq, e) and include neighborhood N (uq, e) in similarity s (e, ei) maximum preceding x activity { ei}；
G3 enables p1As current candidate companion upColleague's similarity, if p1≥pr, then current candidate companion upAs movable e's
Pairing person；
G4, if p1＜ pr, then obtain lower a pair of (u, num), and colleague's similarity of the candidate companion in lower a pair of (u, num)
More than p1When, update p1, and using candidate companion as candidate pairing person；
G5 removes current (u, num) from the membership table, and obtains lower a pair of (u, num), obtain num and | N | compared with
Small value calculates colleague's similarity degree max-thresholds pr, and return to step g3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810294059.4A CN108509610B (en) | 2018-03-30 | 2018-03-30 | Retrieval method and system for activities and companions |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810294059.4A CN108509610B (en) | 2018-03-30 | 2018-03-30 | Retrieval method and system for activities and companions |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108509610A true CN108509610A (en) | 2018-09-07 |
| CN108509610B CN108509610B (en) | 2021-08-06 |
Family
ID=63380124
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810294059.4A Active CN108509610B (en) | 2018-03-30 | 2018-03-30 | Retrieval method and system for activities and companions |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108509610B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582808A (en) * | 2018-11-22 | 2019-04-05 | 北京锐安科技有限公司 | A kind of user information querying method, device, terminal device and storage medium |
| CN112559873A (en) * | 2020-12-21 | 2021-03-26 | 周欢 | User recommendation system based on intelligent education |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060195462A1 (en) * | 2005-02-28 | 2006-08-31 | Yahoo! Inc. | System and method for enhanced media distribution |
| US20130226628A1 (en) * | 2012-02-24 | 2013-08-29 | John D. Scher | Event-centric matching and social networking services |
| CN104915365A (en) * | 2015-02-10 | 2015-09-16 | 腾讯科技（深圳）有限公司 | Activity flow recommending method and device |
| CN106844790A (en) * | 2017-03-22 | 2017-06-13 | 中国人民解放军国防科学技术大学 | A kind of quick Temporal query method for user's group analysis in social networks |
| CN107229704A (en) * | 2017-05-25 | 2017-10-03 | 深圳大学 | A kind of resource description framework querying method and system based on KSP algorithms |
-
2018
- 2018-03-30 CN CN201810294059.4A patent/CN108509610B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060195462A1 (en) * | 2005-02-28 | 2006-08-31 | Yahoo! Inc. | System and method for enhanced media distribution |
| US20130226628A1 (en) * | 2012-02-24 | 2013-08-29 | John D. Scher | Event-centric matching and social networking services |
| CN104915365A (en) * | 2015-02-10 | 2015-09-16 | 腾讯科技（深圳）有限公司 | Activity flow recommending method and device |
| CN106844790A (en) * | 2017-03-22 | 2017-06-13 | 中国人民解放军国防科学技术大学 | A kind of quick Temporal query method for user's group analysis in social networks |
| CN107229704A (en) * | 2017-05-25 | 2017-10-03 | 深圳大学 | A kind of resource description framework querying method and system based on KSP algorithms |
Non-Patent Citations (2)
| Title |
|---|
| DINGMING WU ET AL: "A Density-Based Approach to the Retrieval of Top-K", 《PROCEEDINGS OF THE 25TH ACM INTERNATIONAL ON CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT》 * |
| WENTING TU ET AL: "Activity-Partner Recommendation", 《THE 19TH PACIFIC-ASIA CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109582808A (en) * | 2018-11-22 | 2019-04-05 | 北京锐安科技有限公司 | A kind of user information querying method, device, terminal device and storage medium |
| CN112559873A (en) * | 2020-12-21 | 2021-03-26 | 周欢 | User recommendation system based on intelligent education |
| CN112559873B (en) * | 2020-12-21 | 2021-08-13 | 融易学控股(深圳)有限公司 | User recommendation system based on intelligent education |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108509610B (en) | 2021-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103218436B (en) | A kind of Similar Problems search method and device that merges class of subscriber label | |
| CN108304437A (en) | A kind of automatic question-answering method, device and storage medium | |
| CN108846029B (en) | Information correlation analysis method based on knowledge graph | |
| CN105808590B (en) | Search engine implementation method, searching method and device | |
| KR101543780B1 (en) | System and method for expert search by dynamic profile and social network reliability | |
| EP1995669A1 (en) | Ontology-content-based filtering method for personalized newspapers | |
| Nadimi-Shahraki et al. | Cold-start problem in collaborative recommender systems: Efficient methods based on ask-to-rate technique | |
| CN104298785B (en) | Searching method for public searching resources | |
| CN103823844A (en) | Question forwarding system and question forwarding method on the basis of subjective and objective context and in community question-and-answer service | |
| US20030212663A1 (en) | Neural network feedback for enhancing text search | |
| Wang et al. | Personalized recommendation for new questions in community question answering | |
| CN107526975A (en) | A kind of method based on difference secret protection decision tree | |
| Liu et al. | Double hierarchy hesitant fuzzy linguistic mathematical programming method for MAGDM based on Shapley values and incomplete preference information | |
| Antenucci et al. | Ringtail: Feature Selection For Easier Nowcasting. | |
| CN112989215B (en) | Sparse user behavior data-based knowledge graph enhanced recommendation system | |
| Huang et al. | Improving the relevancy of document search using the multi-term adjacency keyword-order model | |
| CN111144831B (en) | Accurate selection screening system and method suitable for recruitment | |
| Chen et al. | Context-aware ensemble of multifaceted factorization models for recommendation prediction in social networks | |
| CN108509610A (en) | A kind of search method and system of activity and companion | |
| CN108509588B (en) | Lawyer evaluation method and recommendation method based on big data | |
| CN112115379A (en) | Rice variety selection method and device based on knowledge graph | |
| Jamil et al. | A collaborative names recommendation in the Twitter environment based on location | |
| CN113592589B (en) | Textile raw material recommendation method, device and processor | |
| Zaman et al. | Selecting key person of social network using skyline query in MapReduce framework | |
| Zong et al. | Efficiently answering why-not questions on radius-bounded k-core searches |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211230 Address after: 518000 east of the fourth floor of plant 1 (Building 1) of Baode technology R & D and production base, gaoxinyuan, Guanlan street, Longhua new area, Shenzhen, Guangdong Patentee after: Baode network security system (Shenzhen) Co.,Ltd. Address before: 518060 No. 3688 Nanhai Road, Shenzhen, Guangdong, Nanshan District Patentee before: SHENZHEN University |
|
| TR01 | Transfer of patent right |