CN106097110A - A kind of dictionary construction method based on community network and word matched recommend method - Google Patents

Abstract

The invention discloses a kind of dictionary construction method based on community network and word matched recommends method.This recommendation method is: 1) build the individual central network of user x based on community network；2) for user y each in the individual central network of this user x, the similitude of word library between this user x and this user y is calculated；If similitude is more than setting threshold value H1, then calculate cohesion between this user x, user y；If cohesion is more than setting threshold value H2, then determine the score value of word between user x, user y according to the similitude obtaining and cohesion；If score value is more than setting threshold value H3, then the word that this user y remembers in time period T is sent to this user x；3) according to the word sending over, it according to word marking and queuing and is recommended this user x.Invention increases the effect of interaction, and note word efficiency has and is obviously improved.

Description

A kind of dictionary construction method based on community network and word matched recommend method

Technical field

The invention belongs to technical field of network information, relate to a kind of word matched method, more particularly, to one based on The dictionary construction method of community network and word matched recommend method.

Background technology

English, as universal language, is China's language that each field is the most frequently used in addition to Chinese, and English level is Become the important indicator of various examination.The study of word is the basis of English study；How effectively to remember that word always is people The problem extremely paid close attention to.

Existing vocabulary memorization method is to be remembered by vocabulary book or remind memory by software.For example, newly eastern The books such as " six grades of vocabulary root+Associative mnemonics " of side, remember word by word rule；Also has softwares such as " cloud words " at algorithm Have employed human mind rule in design to carry out the prompting user of passive type and remember word.The shortcoming of these methods is that people remember alone list Word, lacks interaction so that note word is uninteresting, inefficiency, it is difficult to adhere to for a long time.

Social network sites (SNS), based on community network, is increasingly becoming people's requisite contact instrument, is a handle There is the platform maintaining the relationship that the people of same interest hobby links together with the friend in life；Good friend in social network sites Relation so that the interaction between user is convenient and direct.

The limitation of existing vocabulary memorization method makes the word that people can not be interested in remember carry out accurate match, base In the word matched method of community network, under conditions of interactive with good friend, word interested for people can be carried out accurately Mate and recommend, allowing vocabulary memorization become more to be rich in enjoyment, with deep impression, also can be easier to adhere to for a long time.

Content of the invention

In view of can not the be interested in people word of memory of tradition vocabulary memorization method carries out the problem of accurate match, this A kind of dictionary construction method based on community network of bright proposition and word matched recommend method.

The method utilizes in community network to be sentenced to cohesion between the demand similarity of word and user between adjacent user Which vocabulary disconnected more mates with the wish of user, thus does the effect that effective word recommends reach again interactive.The method Comprise determining that the demand similitude to word taste or hobby between adjacent user；Between the adjacent user of determination, contact is intimate Degree；It is word scoring according to similitude and cohesion, the word of good friend's memory is recommended user according to scoring.

The technical scheme is that

A kind of dictionary construction method based on community network, the steps include:

1) the individual central network of user x is built based on community network；

2) for user y each in the individual central network of this user x, word library between this user x and this user y is calculated Similitude；If similitude is more than setting threshold value H1, then calculate cohesion between this user x, user y；If cohesion is more than Set threshold value H2, then determine the score value of word between user x, user y according to the similitude obtaining and cohesion；If scoring The word that this user y remembers in time period T more than setting threshold value H3, is then sent to this user x by value；

3) the current word storehouse of the word receiving and this user x is merged and deletes the word of repetition by this user x, Word library to this user x.

A kind of word matched based on community network recommends method, the steps include:

1) the individual central network of user x is built based on community network；

2) for user y each in the individual central network of this user x, word library between this user x and this user y is calculated Similitude；If similitude is more than setting threshold value H1, then calculate cohesion between this user x, user y；If cohesion is more than Set threshold value H2, then determine the score value of word between user x, user y according to the similitude obtaining and cohesion；If scoring The word that this user y remembers in time period T more than setting threshold value H3, is then sent to this user x by value；

3) according to the word sending over, it according to word marking and queuing and is recommended this user x.

Further, the method for the individual central network building user x is: is obtained by community network and is connected with this user x Other individual and these individualities between annexation, constitute the individual central network of this user x.

Further, a graph structure G is used to represent this community network；This graph structure G is two tuple form: G=(V, E)； Wherein, V is a nonempty finite vertex set, each vertex correspondence one user, and E is a limited limit collection.

Further, formula is usedCalculate word library between this user x and this user y Similitude sim (x, y)；Set I represents the token-category set of user's word library, r_x,iRepresent user x to token-category i Scoring, r_y,iRepresent the scoring to token-category i for the user y.

Further, the memory number of times to each word in token-category i for the user is added and obtains this user to this word class The scoring of other i.

Further, actively consult word number of times number of times recommended with this word according to user and be added the note obtaining this word Recall number of times.

Further, the memory number of times of added words in the word library of this user x is designated as 1, the memory of repeated word time Number adds 1.

Further, formula is usedCalculate cohesion between this user x and this user y Fre(x,y)；Wherein, Mes (x, y) the contact quantity between user x, user y, F in expression setting time span in the recent period_xRepresent Good friend's set according to the user x that the individual central network of user x obtains.

Further, the method determining this contact quantity is: counting user x is sent to message number m of user y, user y It is sent to message number n of user x, take the smaller value in m, n as the contact quantity between user x, user y.

Further, according to formula S co=aSim, (x, y) (x y) calculates the score value Sco of word to+bFre；Wherein, a, b For weight coefficient.

Compared with prior art, the positive effect of the present invention is:

By the enforcement of this method, based on user, in social networks, the related information with other users carries out word matched When, can consider that community network adjoins the similarity between user and cohesion simultaneously.In this way, it is recommended that the word giving user is to use What family needed is also the word that its good friend is remembering, and has reached interactive effect simultaneously, and note word efficiency has and is obviously improved.

Brief description

Fig. 1 is that the non-directed graph of the individual central network of user A represents；

Fig. 2 contacts schematic diagram in the recent period for adjoining between user；

Fig. 3 improves figure for contact schematic diagram；

Fig. 4 is word scoring schematic flow sheet；

Fig. 5 is system recommendation schematic flow sheet.

Detailed description of the invention

For the clear detailed description of the invention describing the method accurately, first the term relating in the present invention is said Bright.

Term 1: community network refers to one group of actor and connects their various relations (such as the relation such as friendship, communication) Set；Can be expressed as a graph structure G, figure G is designated as two tuple form: G=(V, E), and wherein V is a nonempty finite summit Collection (i.e. user's set), E is a limited limit collection (i.e. set of relationship), and two users that limit connects or summit are the nothings on summit in V Sequence couple or ordered pair pair.

Term 2: individual central network refers to the actor around a certain particular individual and the collection connecting their various relations Close, obtained other individualities being connected with individuality by community network, and the annexation between these individualities i.e. may make up this Individual individual central network.

Term 3: social network sites refer to based on community network friend and have common interest like user pass through network this The Web site that one carrier associates.

Term 4: user refer to social network sites registration and by certification people be also the action in community network simultaneously Person.

Term 5: friend relation refers to the relation in social network sites between user；Gather if V is user, any subset of V × V The binary crelation that E is referred to as on V, if (x, y) has x ≠ y, then the friend relation that E is referred to as on V, claims x and y for arbitrary element in E There is friend relation.

Term 6: adjoin and refer to exist between user friend relation；If user is x, there is friend relation between y, then claim x and y Adjacent.

Term 7: similarity refers to adjoin the similarity degree to word demand between user in community network.

Term 8: cohesion refers to contact the frequent of (including transmission information, comment etc.) between adjacent user in community network Degree.

Term 9: word scoring refers to according to the comprehensive word score obtaining of cohesion between user and similarity.

Recommending method as a kind of word matched, first system has lexicon, and carries out classified vocabulary as desired, Vocabulary can be such as primary school's English glossary, JEFC vocabulary, Senior High School English vocabulary, College English Test word by grade separation Remittance, College English Test vocabulary etc., can be also automobile, economy, traffic, photography, physics, agricultural by vocabulary according to category classification Deng i.e. system is by classified vocabulary (each word can belong to multiple classification)；And system can record the note list of each user Word situation, including the memory number of times etc. that user is to each word, the memory number of times of word is actively consulted this word by user Number and this word are added by system recommendation number of times and obtain.

For the ease of understanding and realizing the present invention, describe the principle of this recommendation method below with reference to the accompanying drawings in detail.

First, can get the individual central network of any user according to community network, Fig. 1 is the individual central network of user A Figure represents, A, B, C, D, E, F are social network user, and the line between user shows there is friend relation (undirected relation), user A Good friend (i.e. adjoin user) be B, C, D, E, F.It is all to describe in detail with this individuality central network for example below.

Concrete calculation is as follows:

1. calculate similarity, i.e. calculate the similitude to word demand for two users, use cosine similarity.User is current Remember the memory number of times of every class vocabulary as to the scoring of such vocabulary (by the memory to each word in such vocabulary for the user Number of times is added and obtains the memory number of times of every class vocabulary), the scoring of all n class words be counted as one of n-dimensional space vector (as Really user does not remember certain class vocabulary, then the scoring to such vocabulary for the user is set to 0).Assume the scoring of two user x and y to Amount is respectively x and y, and set I represents token-category set, r_x,iAnd r_y,iRepresent the scoring to token-category i for user x and y respectively, Then cosine similarity is

$Sim(x,y)=cos(x,y)=\frac{\underset{i\∈I}{\mathrm{\Σ}}{r}_{x,i}{r}_{y,i}}{\sqrt{\underset{i\∈I}{\mathrm{\Σ}}{r}_{x,i}^2}\sqrt{\underset{i\∈I}{\mathrm{\Σ}}{r}_{y,i}^2}}---\left(1\right)$

Wherein, 0≤Sim (x, y)≤1.

2. calculate cohesion, i.e. calculate the frequent degree of two user's contacts.If Fig. 3 is the individual central network of user A Contact figure, the direction of arrow is that user issues the message such as the information of good friend, evaluation in the recent period, and the numeral on arrow represents that recent user sends out To the message sum of good friend.Generally between good friend, messaging more expressions both sides more get close to, but if only one-side Message sends and is then probably harassing and wrecking information without replying, and this message number can not show that two people's gets close to degree, so Here take the less quantity contacting as two people of interaction message number between both sides, then Fig. 3 is improved to Fig. 4.Assume (x, y) represents recent x, contacts quantity, F between y for two user x and y, Mes_xRepresent in the good friend's set i.e. individuality of x of user x The user's set being connected with x in heart net, then contacting frequent degree is

$Fre(x,y)=\frac{Mes(x,y)}{\underset{f\∈F_x}{\mathrm{\Σ}}Mes(x,f)}---\left(2\right)$

Wherein 0≤Mes (x, y)≤1.Cohesion Fre (A, B)=0.43 such as user A and user B in Fig. 3.

3. calculate word scoring, i.e. calculate the comprehensive score of similarity and cohesion.Then word comprehensively must be divided into

Sco=aSim (x, y)+bFre (x, y) (3)

Wherein, the value of a+b=1, a and b represents the weight that similarity and cohesion account for, and arranges according to actual conditions；Can root According to actual conditions be similarity and cohesion is respectively provided with corresponding threshold value, calculates the user only considering more than threshold value.

Fig. 4 is word scoring flow chart, specifically comprises the following steps that

Step 1: be calculated similarity with formula (1)；

Step 2: judge whether similarity is less than threshold value, if it is, be set to 0 and terminate scoring by word score value；

Step 3: be calculated cohesion with formula (2)；

Step 4: judge whether cohesion is less than threshold value, if it is, be set to 0 and terminate scoring by word score value；

Step 5: be calculated word score value with formula (3).

Fig. 5 is system word recommended flowsheet figure, specifically comprises the following steps that

Step 1: judge whether the total words that the good friend of user records a demerit is 0, if it is, terminate to recommend.

Step 2: the word score value of step calculating the remembered word of each good friend of user of flow chart of marking according to word, and shape Become dictionary to be recommended；

Step 3: the word that word score value is 0 is directly deleted, and updates dictionary to be recommended；

Step 4: judge whether dictionary to be recommended is empty, if it is, terminate to recommend.

Step 5: merge the word repeating to recommend；It because society's figure is to have ring figure, is likely to same between good friend Word repeats to recommend, and same word merged into one by this step, and the list after giving merging by this word score value soprano Word；Update dictionary to be recommended；

Step 6: the word in dictionary to be recommended is ranked up by the score value according to each word according to size, is pushed away Recommend word list.

Step 7: when user logs in social network sites for the first time, word list is recommended user in every day.

Therefore embodiment described above only have expressed the several embodiments of the present invention, describes more detailed, but can not be It is interpreted as the restriction of the scope of the claims of the present invention.It should be pointed out that, for the technical staff of areas of information technology, without departing from this On the premise of inventive concept, can also make different distortion and improvement, these broadly fall into protection scope of the present invention.

Claims (10)

1. the dictionary construction method based on community network, the steps include:

1) the individual central network of user x is built based on community network；

2) for user y each in the individual central network of this user x, the similar of word library between this user x to this user y is calculated Property；If similitude is more than setting threshold value H1, then calculate cohesion between this user x, user y；If cohesion is more than setting Threshold value H2, then determine the score value of word between user x, user y according to the similitude obtaining and cohesion；If score value is big In setting threshold value H3, then the word that this user y remembers in time period T is sent to this user x；

3) the current word storehouse of the word receiving and this user x is merged and deletes the word of repetition by this user x, is somebody's turn to do The word library of user x.

2. recommend a method based on the word matched of community network, the steps include:

1) the individual central network of user x is built based on community network；

2) for user y each in the individual central network of this user x, the similar of word library between this user x to this user y is calculated Property；If similitude is more than setting threshold value H1, then calculate cohesion between this user x, user y；If cohesion is more than setting Threshold value H2, then determine the score value of word between user x, user y according to the similitude obtaining and cohesion；If score value is big In setting threshold value H3, then the word that this user y remembers in time period T is sent to this user x；

3) according to the word sending over, it according to word marking and queuing and is recommended this user x.

3. method as claimed in claim 1 or 2, it is characterised in that the method for the individual central network building user x is: pass through Community network obtains the annexation between other individualities being connected with this user x and these individualities, constitutes the individual of this user x Body central network.

4. method as claimed in claim 3, it is characterised in that use a graph structure G to represent this community network；This graph structure G It is two tuple form: G=(V, E)；Wherein, V is a nonempty finite vertex set, and each vertex correspondence one user, E is one to be had Limit limit collection.

5. method as claimed in claim 1 or 2, it is characterised in that use formulaCalculate Between this user x and this user y word library similitude sim (x, y)；Set I represents the token-category set of user's word library, r_x,iRepresent the scoring to token-category i for the user x, r_y,iRepresent the scoring to token-category i for the user y.

6. method as claimed in claim 5, it is characterised in that the memory number of times phase to each word in token-category i for the user Add the scoring obtaining this user to this token-category i.

7. method as claimed in claim 6, it is characterised in that actively consult word number of times according to user recommended with this word Number of times is added the memory number of times obtaining this word；The memory number of times of added words in the word library of this user x is designated as 1, repeats The memory number of times of word adds 1.

8. method as claimed in claim 1 or 2, it is characterised in that use formulaCalculating should Cohesion Fre between user x and this user y (x, y)；Wherein, (x y) represents and sets user x, use in time span in the recent period Mes Contact quantity between the y of family, F_xRepresent good friend's set of the user x obtaining according to the individual central network of user x.

9. method as claimed in claim 8, it is characterised in that the method determining this contact quantity is: counting user x is sent to Message number m of user y, user y are sent to message number n of user x, take smaller value in m, n as user x, user y it Between contact quantity.

10. method as claimed in claim 1 or 2, it is characterised in that (x, y) (x y) counts+bFre according to formula S co=aSim Calculate the score value Sco of word；Wherein, a, b are weight coefficient.