CN106055616A - Friend recommendation method for social networking website based on named entities - Google Patents

Abstract

The invention provides a friend recommendation method for social networking website based on named entities. The method comprises following steps: establishing an entity list ConEntity (ui) composed of entities during users' speaking and entities of followers in speaking, establishing an entity list InfEntity (ui) for personal information and sequencing entities of users' preference based on obtained entity list ConEntity (ui) and the entity list InfEntity (ui) to obtain an entity list for preference sequences. Through similarity between entity lists of users, friends are recommended for users with similar interests in the social networking website.

Description

A kind of friends in social networking sites based on name entity recommends method

Technical field

The present invention relates to natural language processing technique field, more particularly, to a kind of social network based on name entity Stand friend recommendation method.

Background technology

In recent years, along with the fast development of the Internet, large-scale knowledge base is (such as wikipedia, DBpedia, Baidu hundred Section etc.) occur in the Internet and be rapidly developed.These knowledge bases are rich in about the language between attribute and the entity of limbs Justice relations etc., their appearance makes user can obtain relevant information easily.Therefore, the technology about name entity also exists Develop rapidly and be applied to the every field of the Internet.

So-called name entity, it is simply that name, mechanism's name, place name and other all entities with entitled mark.Wider General entity also includes numeral, date, currency, address etc..Existing technology relates generally to name the identification of entity, and name is real The link of body and disambiguation and the relation excavation field of name entity, and the most ripe.The identification of name entity refers to, From given a word or an article, find the word referring to name entity, we term it entity censures item, In " Jordon is famous basket baller ", it is " Jordon " that our entity to be identified censures item.The link of name entity is Refer to that we to censure item and the entity (showing as a page in wikipedia) in certain knowledge base determined identify It is chained up reaching the purpose of disambiguation.Such as, " Jordon is famous basket baller " and " Jordon is that U.S.'s Berkeley is big The professor of the research machine learning learned " in two identical denotion items " Jordon " point to is diverse two entities.And The contact between two entities is mainly looked in the excavation of entity relationship, " Jordon " and " Berkeley during as above an example is waited University " relation of university is " A teaches in B ".

Existing social networks friend recommendation method mainly has based on customer relationship with based on label and the big class of content two: base Recommendation in customer relationship mainly has the common friend recommending user, it is recommended that the good friend of good friend, it is recommended that the follower etc. of follower Deng, and work in coordination with based on interest, if i.e. user A is similar with user B vermicelli crowd, and user to be recommended has paid close attention to user A, that User B is also recommended this user；Expansion is come, and can find the community structure residing for user based on customer relationship figure, will Other in same corporations are not paid close attention to member and are recommended this user.Method based on label and content is mainly according to user's Geographical location information, educational work information, or user-defined individual's label, it is recommended that other similar users.Base The recommended intensity of the method for plinth is identical, does not has the difference according to the probability size becoming good friend and has different recommendations Dynamics.And other traditional methods or the most sufficiently utilize speech information and the label information of user, or by user's Good friend is confined in certain interest circle or friend circle, it is impossible to fully find the potential targets of interest of user.

Summary of the invention

The present invention provides a kind of friends in social networking sites based on name entity to recommend method, it is achieved the use in social networks Other users of similar preference are recommended at family.

In order to reach above-mentioned technique effect, technical scheme is as follows:

A kind of friends in social networking sites based on name entity recommends method, comprises the following steps:

S1: to the user u in candidate user collection U_iAnd the speech text of follower carries out pretreatment and is named reality The identification of body and connection, draw by user u_iEntity in speech and entity row of the entity composition in its follower speech Table ConEntity (u_i)；

S2: to user u_iCarry out personal information extraction and be named identification and the connection of entity, drawing about user u_iIndividual List of entities InfEntity (the u of people's information_i)；

S3: according to the list of entities ConEntity (u obtained_i) and InfEntity (u_i) to user u_iThe entity of preference It is ranked up, obtains the list of entities UserEntity (u of a preference sequence_i)；

S4: utilize the UserEntity (u of each user_i) carrying out similarity comparison, the user choosing highest similarity enters Row is mutually recommended.

Preferably, to user u in described step S1_iAnd the speech text of follower carries out the mode of pretreatment and includes point Word, stop words filter.

Preferably, the userspersonal information in described step S2 includes geographical location information, hobby label.

Further, the detailed process of described step S3 is as follows:

K is made to represent the significance level that user mentions for entity e:

K=count_ConEntity(e)+1.2count_InfEntity(e)

Wherein, count_ConEntity(e) presentation-entity e occurrence number in set ConEntity (U), count_InfEntity(e) presentation-entity e occurrence number in set ConEntity (U)；

User's preference value to entity:

$Perfer(u_i,e)=K\·\left(Sim\right(u_i,e)+\mathrm{\α}\·\frac{count\left(Entity\right(u_i)\∩Entity\left(e\right))}{\min \left(count\right(Entity\left(u_i\right),Entity\left(e\right)\left)\right)}+\mathrm{\β}\·\frac{1}{\sqrt{count\left(in\right(e)+1)}})$

Wherein, Sim (u_i, e) represent user u_iWith the text similarity of entity e, Entity (e) the presentation-entity page is mentioned The entity sets in addition to e, Entity (u_i)=ConEntity (u_i)∩InfEntity(u_i) represent that user two is relevant real The set of body, in (e) represents the number of the entity pointing to entity e,The popularity of presentation-entity, α and β is Weight parameter.

Further, according to user, the preference value of entity is obtained user u_iThe list of entities of preference sequence, chooses it Middle top n entity forms list of entities UserEntity (u_i), during less than N entity, room goes out fills with null.

Further, the detailed process of step S4 is as follows:

The distance of two inter-entity of calculating:

$\mathrm{Re}l(e_1,e_2)=1-\frac{log\left(max\right(\left|g\right(e_i)|,|g\left(e_j\right)\left|\right))-log\left(max\right(\left|g\right(e_i)|\∩\left|g\left(e_j\right)\right|\left)\right)}{log\left(max\right(\left|Total\right|))-log\left(\min \right(\left|g\right(e_i)|,\left|g\left(e_j\right)\right|\left)\right)}$

Wherein, Total is the quantity of all entities in knowledge base, and g (e) has link to point to entity e in knowledge base The set of entity, for user u₁With user u₂, have list of entities UserEntity (u₁) and UserEntity (u₂), e_i∈ UserEntity(u₁), e_j∈UserEntity(u₂), i, j ∈ [1, N], e_iWith user u₂List similarity:

$\mathrm{Re}f(e_i,UserEntity(u_2\left)\right)=\underset{e_j\∈UserEntity\left(u_2\right),j=1}{\overset{N}{\Σ}}\mathrm{Re}l(e_i,e_j)$

User u₁With user u₂Between similarity be:

$R(u_1,u_2)=\underset{e_i\∈UserEntity\left(u_1\right),i=1}{\overset{N}{\Σ}}\frac{\mathrm{Re}l(e_i,UserEntity(u_2))}{log(1+i)}.$

Compared with prior art, technical solution of the present invention provides the benefit that:

The present invention sets up the list of entities that the entity during the entity in being made a speech is made a speech with its follower forms by user ConEntity(u_i), set up the list of entities InfEntity (u of userspersonal information_i), and according to the list of entities obtained ConEntity(u_i) and InfEntity (u_i) entity of the preference of user is ranked up obtaining the entity row of preference sequence Table, carries out, to the user in social networks, the friend recommendation that interest is similar by the similarity degree of list of entities between user.

Accompanying drawing explanation

Fig. 1 is the inventive method flow chart.

Detailed description of the invention

Accompanying drawing being merely cited for property explanation, it is impossible to be interpreted as the restriction to this patent；

In order to the present embodiment is more preferably described, some parts of accompanying drawing have omission, zoom in or out, and do not represent actual product Size；

To those skilled in the art, in accompanying drawing, some known features and explanation thereof may be omitted is to be appreciated that 's.

With embodiment, technical scheme is described further below in conjunction with the accompanying drawings.

Embodiment 1

As it is shown in figure 1, a kind of friends in social networking sites based on name entity recommends method, comprise the following steps:

S1: to the user u in candidate user collection U_iAnd the speech text of follower carries out pretreatment and is named reality The identification of body and connection, draw by user u_iEntity in speech and entity row of the entity composition in its follower speech Table ConEntity (u_i)；

S2: to user u_iCarry out personal information extraction and be named identification and the connection of entity, drawing about user u_iIndividual List of entities InfEntity (the u of people's information_i)；

S3: according to the list of entities ConEntity (u obtained_i) and InfEntity (u_i) to user u_iThe entity of preference It is ranked up, obtains the list of entities UserEntity (u of a preference sequence_i)；

S4: utilize the UserEntity (u of each user_i) carrying out similarity comparison, the user choosing highest similarity enters Row is mutually recommended.

To user u in step S1_iAnd the speech text of follower carries out the mode of pretreatment and includes participle, stop words mistake Filter.

Userspersonal information in step S2 includes geographical location information, hobby label.

The detailed process of step S3 is as follows:

K is made to represent the significance level that user mentions for entity e:

K=count_ConEntity(e)+1.2count_InfEntity(e)

Wherein, count_ConEntity(e) presentation-entity e occurrence number in set ConEntity (U), count_InfEntity(e) presentation-entity e occurrence number in set ConEntity (U)；

User's preference value to entity:

$Perfer(u_i,e)=K\·\left(Sim\right(u_i,e)+\mathrm{\α}\·\frac{count\left(Entity\right(u_i)\∩Entity\left(e\right))}{\min \left(count\right(Entity\left(u_i\right),Entity\left(e\right)\left)\right)}+\mathrm{\β}\·\frac{1}{\sqrt{count\left(in\right(e)+1)}})$

Wherein, Sim (u_i, e) represent user u_iWith the text similarity of entity e, Entity (e) the presentation-entity page is mentioned The entity sets in addition to e, Entity (u_i)=ConEntity (u_i)∩InfEntity(u_i) represent that user two is relevant real The set of body, in (e) represents the number of the entity pointing to entity e,The popularity of presentation-entity, α and β is Weight parameter, can be obtained by training set training.

According to user, the preference value of entity is obtained user u_iThe list of entities of preference sequence, chooses wherein top n real Body forms list of entities UserEntity (u_i), during less than N entity, room goes out fills with null.In the present embodiment Europe, N takes 30, During less than 30 entity, room goes out fills with null.

The detailed process of step S4 is as follows:

The distance of two inter-entity of calculating:

$\mathrm{Re}l(e_1,e_2)=1-\frac{log\left(max\right(\left|g\right(e_i)|,|g\left(e_j\right)\left|\right))-log\left(max\right(\left|g\right(e_i)|\∩\left|g\left(e_j\right)\right|\left)\right)}{log\left(max\right(\left|Total\right|))-log\left(\min \right(\left|g\right(e_i)|,\left|g\left(e_j\right)\right|\left)\right)}$

Wherein, Total is the quantity of all entities in knowledge base, and g (e) has link to point to entity e in knowledge base The set of entity, for user u₁With user u₂, have list of entities UserEntity (u₁) and UserEntity (u₂), e_i∈ UserEntity(u₁), e_j∈UserEntity(u₂), i, j ∈ [1, N], e_iWith user u₂List similarity:

$\mathrm{Re}f(e_i,UserEntity(u_2\left)\right)=\underset{e_j\∈UserEntity\left(u_2\right),j=1}{\overset{30}{\Σ}}\mathrm{Re}l(e_i,e_j)$

User u₁With user u₂Between similarity be:

$R(u_1,u_2)=\underset{e_i\∈UserEntity\left(u_1\right),i=1}{\overset{30}{\Σ}}\frac{\mathrm{Re}l(e_i,UserEntity(u_2))}{log(1+i)}.$

Obtain and can mutually recommend between the user of highest similarity.Such as user u₁, calculate respectively he with The similarity of all users in other candidate user set, then chooses front n user and carries out friend recommendation or interested User pay close attention to recommendation.

This method sets up the entity row that the entity during the entity in being made a speech is made a speech with its follower forms by user ui Table ConEntity (ui), sets up list of entities InfEntity (ui) of userspersonal information, and according to the list of entities obtained ConEntity (ui) and InfEntity (ui) is ranked up obtaining the entity row of preference sequence to the entity of the preference of user Table, carries out, to the user in social networks, the friend recommendation that interest is similar by the similarity degree of list of entities between user.

The corresponding same or analogous parts of same or analogous label；

Described in accompanying drawing, position relationship is used for the explanation of being merely cited for property, it is impossible to be interpreted as the restriction to this patent；

Obviously, the above embodiment of the present invention is only for clearly demonstrating example of the present invention, and is not right The restriction of embodiments of the present invention.For those of ordinary skill in the field, the most also may be used To make other changes in different forms.Here without also cannot all of embodiment be given exhaustive.All at this Any amendment, equivalent and the improvement etc. made within the spirit of invention and principle, should be included in the claims in the present invention Protection domain within.

Claims (6)

1. a friends in social networking sites based on name entity recommends method, it is characterised in that comprise the following steps:

S1: to the user u in candidate user collection U_iAnd the speech text of follower carries out pretreatment and is named the knowledge of entity And connect, do not draw by user u_iEntity in speech and a list of entities of the entity composition in its follower speech ConEntity(u_i)；

S2: to user u_iCarry out personal information extraction and be named identification and the connection of entity, drawing about user u_iIndividual's letter List of entities InfEntity (the u of breath_i)；

S3: according to the list of entities ConEntity (u obtained_i) and InfEntity (u_i) to user u_iThe entity of preference carry out Sequence, obtains the list of entities UserEntity (u of a preference sequence_i)；

S4: utilize the UserEntity (u of each user_i) carrying out similarity comparison, the user choosing highest similarity is carried out mutually Recommend.

Friends in social networking sites based on name entity the most according to claim 1 recommends method, it is characterised in that described step To user u in rapid S1_iAnd the speech text of follower carries out the mode of pretreatment and includes that participle, stop words filter.

Friends in social networking sites based on name entity the most according to claim 1 recommends method, it is characterised in that described step Userspersonal information in rapid S2 includes geographical location information, hobby label.

Friends in social networking sites based on name entity the most according to claim 1 recommends method, it is characterised in that described step The detailed process of rapid S3 is as follows:

K is made to represent the significance level that user mentions for entity e:

K=count_ConEntity(e)+1.2count_InfEntity(e)

Wherein, count_ConEntity(e) presentation-entity e occurrence number in set ConEntity (U), count_InfEntity(e) Presentation-entity e occurrence number in set ConEntity (U)；

User's preference value to entity:

$Perfer(u_i,e)=K\·\left(Sim\right(u_i,e)+\mathrm{\α}\·\frac{count\left(Entity\right(u_i)\∩Entity(e\left)\right)}{\min \left(count\right(Entity\left(u_i\right),Entity\left(e\right)\left)\right)}+\mathrm{\β}\·\frac{1}{\sqrt{count\left(in\right(e)+1)}})$

Wherein, Sim (u_i, e) represent user u_iWith the text similarity of entity e, Entity (e) the presentation-entity page mention except e Outside entity sets, Entity (u_i)=ConEntity (u_i)∩InfEntity(u_i) represent two related entities of user Set, in (e) represents the number of the entity pointing to entity e,The popularity of presentation-entity, α and β is weight Parameter.

The most according to claim 4 based on name entity friends in social networking sites recommend method, it is characterised in that according to Family obtains user u to the preference value of entity_iThe list of entities of preference sequence, chooses wherein top n entity and forms list of entities UserEntity(u_i), during less than N entity, room goes out fills with null.

Friends in social networking sites based on name entity the most according to claim 5 recommends method, it is characterised in that step S4 Detailed process as follows:

The distance of two inter-entity of calculating:

$\mathrm{Re}l(e_1,e_2)=1-\frac{log\left(max\right(\left|g\left(e_i\right)\right|,\left|g\left(e_j\right)\right|\left)\right)-log\left(max\right(\left|g\left(e_i\right)\right|\∩\left|g\left(e_j\right)\right|\left)\right)}{log\left(max\right(\left|Total\right|\left)\right)-\log \left(\min \right(\left|g\left(e_i\right)\right|,\left|g\left(e_j\right)\right|\left)\right)}$

Wherein, Total is the quantity of all entities in knowledge base, and g (e) is the entity having link to point to entity e in knowledge base Set, for user u₁With user u₂, have list of entities UserEntity (u₁) and UserEntity (u₂), e_i∈ UserEntity(u₁), e_j∈UserEntity(u₂), i, j ∈ [1, N], e_iWith user u₂List similarity:

$\mathrm{Re}f(e_i,UserEntity(u_2\left)\right)=\underset{e_j\∈UserEntity\left(u_2\right),j=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{Re}l(e_i,e_j)$

User u₁With user u₂Between similarity be:

$R(u_1,u_2)=\underset{e_i\∈UserEntity\left(u_1\right),i=1}{\overset{N}{\mathrm{\Σ}}}\frac{\mathrm{Re}l(e_i,UserEntity(u_2\left)\right)}{log(1+i)}.$