CN101061713A - Method and device for updating user profile - Google Patents

Abstract

To form a mark structure capable of being utilized as an alignment mark in subsequent processes on a semiconductor film during the same exposure process in a process for obtaining a semiconductor of a large particle size crystal phase from a semiconductor film.The optical modulation element 3 has a light intensity modulation structure SP forming a light intensity distribution for crystallization by modulating light, and a mark formation structure MK provided integrally with or independently from the light intensity modulation structure in order to form a light intensity distribution including a predetermined pattern by modulating light and to indicate a predetermined position of a crystallization region. According to the optical modulation element, a crystal can be grown in a predetermined direction from a crystalline nucleus formed at an arbitrary position of a semiconductor film deposited on an insulation substrate with a predetermined thickness, and an alignment mark AM can be formed at an arbitrary position of the semiconductor film in the same process.

Description

A kind of method and apparatus that upgrades files on each of customers

Technical background

The present invention relates to a kind of information recommendation system, relate in particular to the method and apparatus of files on each of customers in a kind of lastest imformation commending system.

Along with the development of modern communication technology, people can obtain bulk information at any time.Yet the rapid increase of amount of information makes people often at a loss as to what to do, and people have pressed for a kind of instrument and can help them to find real interested content, i.e. Ge Xinghua information recommendation system fast.

In order to satisfy the user interest that constantly changes, the files on each of customers in the information recommendation system (UserProfile) also needs constantly to change.So, how, dynamically revise the files on each of customers in the commending system according to user's interest (hobby), to recommend the real interested content of user to the user, become the current problem that needs solution.

At present usually watch the manner of a specific program to revise in the files on each of customers user to the fancy grade and the weight of each content characteristic of this specific program according to the user.Described user's manner is meant that the user watches the time span of a specific program, program with certain content feature watched number of times and deletion number of times.

But each actor names that comprises in the described content characteristic dactylus order (Fan Bingbing, Pueraria lobota excellent), program category (literary film, romance movie, horror film etc.), director (Zhang Yimou, Feng Xiaogang etc.) etc., these content characteristics can be from information sources such as broadcasting, TV or internet, the most representatively sends to the user by electronic program guide of digital television (EPG) with program exactly.

Yet, the user watch the number of times that comprises the certain content feature and deletion number of times only representative of consumer whether watch, can not react the user exactly and whether like these content characteristics.For example, the user only is through a program that comprises the certain content feature because changing channel, representative of consumer is not just liked this content characteristic, thereby be considered the user and watched this content characteristic once, and then as the foundation of revising files on each of customers, this way obviously can not reflect the interestization that the user is real exactly.

Moreover after the user need watch one section content usually, whether just might confirming controls oneself liked this specific program.If but after having seen, the user also dislikes this program, but system thinks that still the user watched this program, and according to content corresponding feature in this program, the user can not be reflected user's interests change so exactly to the fancy grade and the weight of described corresponding contents feature in the correction files on each of customers.

Equally, if only watch the ratio of the total length of the time span of a specific program and this program to revise that the user neither be very accurate to the fancy grade and the weight of each content characteristic in this specific program in the files on each of customers according to the user.Original as the predetermined broadcast total time length of a specific program is not very long just, the user just accounted for the time of trying total time in most or all after, just find oneself not to like this program, utilize above-mentioned watch total time length and the ratio of this program total length revise fancy grade and the weight of user to each content characteristic in this specific program, just very the accurate response user to the interests change of described content characteristic.

For another example, the user (accompanies friend to watch because of boring, or others watches etc.) and watch a specific program, only therefore go to revise fancy grade and the weight of user, real interests change that can not the accurate response user to this each content characteristic of specific program as normal condition.

In sum, at present only watching a specific program, viewing time length with the user or not watching the manner of a specific program serves as can not reflect user's real interests change according to revising in the files on each of customers user very comprehensively, exactly to the fancy grade and the weight of each content characteristic in the described specific program.

Therefore, need provide a kind of method, device and corresponding information recommendation system thereof that upgrades files on each of customers, to revise files on each of customers more comprehensively, exactly.

Summary of the invention

One of purpose of the present invention is in order to revise files on each of customers more comprehensively, exactly, and a kind of method, device and the corresponding information recommendation system thereof that upgrades files on each of customers that provides.

The method of renewal files on each of customers of the present invention, this files on each of customers comprise the fancy grade of a user at least one content characteristic, and the method comprising the steps of: (a). and monitor this user's behavior, the behavior is at a program in broadcast; (b). according to this user behavior, obtain the interest-degree of this user to this program, this interest-degree is at content characteristic that this program presetted; (c). according to the fancy grade of content characteristic corresponding in the described files on each of customers, this interest-degree is correspondingly adjusted; (d). according to adjusted interest-degree, revise the fancy grade of content characteristic corresponding in the described files on each of customers at the content characteristic that this program presetted.

Described files on each of customers comprises the weight of user to described at least one content characteristic, and described method also comprises step: according to adjusted interest-degree at the content characteristic that this program presetted, revise the weight of the content characteristic of described files on each of customers correspondence.

One embodiment of the present of invention are if the fancy grade of corresponding content characteristic shows that the user loses interest in to this content characteristic in described user's case shelves, then to adjust described interest-degree, to reduce the influence of this interest-degree to described files on each of customers.

An alternative embodiment of the invention is, and is interested in this content characteristic if described fancy grade shows the user, then adjusts described interest-degree, to increase the influence of this interest-degree to described files on each of customers.

One of method of renewal files on each of customers of the present invention, be after the total time length ratio of time span of watching a specific program according to the user and the broadcast of this program predetermination obtains the interest-degree of user to this program, again user in this interest-degree and the files on each of customers is compared the fancy grade of each content characteristic in this program or other historical record (as the number of times of the specific program that comprises one or more content characteristics as described in watching or deletion number of times etc.), and come described interest-degree is adjusted, thereby obtain user's interest-degree more accurately according to comparative result.

For example, if or the number of times watched very little to the fancy grade of a content characteristic seldom in the original files on each of customers, then reduce of the influence of described interest-degree to described fancy grade; If it is originally a lot of in the files on each of customers to the fancy grade of content characteristic number of times very big or that watch, then can not reduce (can the very little situation of aforementioned relatively fancy grade, reduce among a small circle) yet even strengthen of the influence of described interest-degree described fancy grade.

Therefore, use method of the present invention to revise files on each of customers, reduced under some special scenes, careless or change platform and watched certain program or accompanied friend to watch certain program etc. as the user, go to revise the possibility of files on each of customers by normal condition, upgrade files on each of customers thereby change according to user interest more accurately.

The device of renewal files on each of customers of the present invention, this files on each of customers comprises the fancy grade of a user at least one content characteristic, and this device comprises a user interaction means, an interests change analytical equipment, an interests change adjusting device and a files on each of customers correcting device.Wherein this user interaction means is used to monitor this user's behavior, and the behavior is at a program in broadcast.This interests change analytical equipment obtains the interest-degree of this user to this program according to this user behavior, and this interest-degree is at content characteristic that this program presetted.This interests change adjusting device is used for the fancy grade according to the corresponding content characteristic of described files on each of customers, and this interest-degree is correspondingly adjusted.This files on each of customers correcting device is used for according to adjusted interest-degree at the content characteristic that this program presetted, and revises the fancy grade of content characteristic corresponding in the described files on each of customers.

Wherein said files on each of customers comprises the weight of user to described at least one content characteristic, wherein said files on each of customers correcting device also is used for revising the type weight of the content characteristic of described files on each of customers correspondence according to adjusted interest-degree at the content characteristic that this program presetted.

One embodiment of the present of invention are, this interests change adjusting device also is used for obtaining the fancy grade of content characteristic corresponding in the described files on each of customers, if this fancy grade shows the user this content characteristic is lost interest in, then adjust described interest-degree, to reduce of the influence of this interest-degree to described files on each of customers.

An alternative embodiment of the invention is, this interests change adjusting device also is used for obtaining the fancy grade of the corresponding content characteristic of described files on each of customers, if it is interested in this content characteristic that described fancy grade shows the user, then adjust described interest-degree, to increase of the influence of this interest-degree to described files on each of customers.

Therefore, use device of the present invention to revise files on each of customers, reduced under some special scenes, careless or change platform and watched certain program or accompanied friend to watch certain program etc. as the user, go to revise the possibility of files on each of customers by normal condition, upgrade files on each of customers thereby change according to user interest more accurately.

Information recommendation system of the present invention comprises a performance receiving apparatus, a files on each of customers management devices, a screening plant, a user interaction means, an interests change analytical equipment, an interests change adjusting device and a files on each of customers correcting device.Wherein this performance receiving apparatus is used for program receiving information.This files on each of customers management devices is used for the stored user archives, and this files on each of customers comprises the fancy grade of a user at least one content characteristic.This screening plant is used for selecting the programme information that the user may like according to described files on each of customers from described programme information, gives the user to recommend this programme information of selecting.This user interaction means is used to monitor this user's behavior, and the behavior is at institute's recommend programs information.This interests change analytical equipment is used for according to this user behavior, obtains the interest-degree of this user to this program, and this interest-degree is at content characteristic that this program presetted.This interests change adjusting device is used for the fancy grade according to the corresponding content characteristic of described files on each of customers, and this interest-degree is correspondingly adjusted.This files on each of customers correcting device is used for according to adjusted interest-degree at the content characteristic that this program presetted, and revises the fancy grade of content characteristic corresponding in the described files on each of customers.

The method of renewal files on each of customers of the present invention, device and corresponding information recommendation system thereof, watch in the manner of a specific program and the original subscriber's archives user to the fancy grade and the weight of each content characteristic in this specific program in conjunction with the user, revise fancy grade and the weight of user to described each content characteristic, thereby follow the tracks of user's interests change more comprehensively, exactly, revise fancy grade and the weight of user content characteristic.

The present invention has also reduced under some special scenes, careless or change platform and watched certain program or accompanied friend to watch certain program etc. as the user, go to revise the possibility of files on each of customers by normal condition, upgrade files on each of customers thereby change according to user interest more accurately.

By following description and the claim that reference is carried out in conjunction with the accompanying drawings, other purpose of the present invention and achievement will be conspicuous, and the present invention is also had more comprehensively understanding.

Description of drawings

The present invention carries out detailed explanation by the mode of example with reference to accompanying drawing, wherein:

Fig. 1 is the structural representation of an information recommendation system according to an embodiment of the invention;

Fig. 2 is flow chart that upgrades the files on each of customers method according to an embodiment of the invention;

Fig. 3 is another flow chart that upgrades files on each of customers according to an embodiment of the invention;

Fig. 4 is the fuzzy set curve chart of input variable e1 among Fig. 3;

Fig. 5 is the fuzzy set curve chart of input variable e2 among Fig. 3;

Fig. 6 is output variable α among Fig. 3 _IjThe fuzzy set curve chart.

In all accompanying drawings, identical reference number is represented similar or identical feature and function.

Embodiment

Fig. 1 is the structural representation of an information recommendation system according to an embodiment of the invention.

This system 100 comprises user interaction means 103, interests change analytical equipment 104, interests change adjusting device 105 and a files on each of customers correcting device 106.Wherein said files on each of customers comprises fancy grade and the weight of a user at least one content characteristic.

Wherein said content characteristic is meant each actor names of comprising in the program (Fan Bingbing, Pueraria lobota excellent), program category (literary film, romance movie, horror film etc.), director (Zhang Yimou, Feng Xiaogang etc.) etc.These content characteristics can be from any information sources such as broadcasting, TV or internet, the most representatively sends to the user by electronic program guide of digital television (EPG) with program exactly.

Content characteristic in the files on each of customers can be one, as has only certain performer.Certainly, the content characteristic in the files on each of customers also can have a plurality of, and at this moment, corresponding recommendation results will be more accurate.

Described fancy grade is meant the sensation of user to each content characteristic, can preset a number range by supplier, as [50 ,+50].

Described weight is meant the user when selecting program, various dissimilar content characteristics, as, performer, director, program category etc. are to the influence of its selection result.What also we can say the user when the program of selecting oneself to like, be standard normally with, promptly still directs according to performer, program category and wait the program of selecting it to like.Wherein all performers' weight can be the same, and the weight of all program categories can be the same, and all directors' weight also can be the same.Described weight also can preset a number range by supplier, as [0,100].

Weight in the described files on each of customers and fancy grade, can be described as the historical record that the user watches program, when the user watches program, other history information can be arranged also, watch the number of times of a program that comprises the certain content feature and deletion number of times etc. as the user.

This user interaction means 103 is used to monitor a user's behavior, and the behavior is at a program in broadcast.This user interaction means 103 can receive the feedback information of user to its program of watching as user and the mutual bridge of commending system, also can present recommendation information and tabulate and select the program that need watch for the user to the user.This feedback information comprises user's behavior.

The described recommendation information available following tabulation lattice 1 of tabulating are represented:

Form 1

Channel	Reproduction time	Title	Possible interest-degree
				HNTV	September 18 15:30	Empty mirror	60 (very interesting)
One in central authorities	September 18 19:30	Tell the truth	45 (being interested in)
				Six in central authorities	September 18 21:30	OK a karaoke club is the bar dog	20 (more interesting)

This interests change analytical equipment 104 is used for according to above-mentioned user behavior, obtains the interest-degree of this user to described program, and this interest-degree is at content characteristic that this program presetted.

Described interest-degree can be expressed as:

WD wherein _iThe time span that the expression user watches a specific program, θ represents a predetermined threshold value, can be provided in advance by supplier, generally can be 0.5RD _i, RD _iThe total time length of representing the predetermined broadcast of this specific program.

This interests change adjusting device 105 is used for the fancy grade according to the corresponding content characteristic of described files on each of customers, and this interest-degree is correspondingly adjusted.An available different coefficient comes this interest-degree is adjusted, and the magnitude range of this coefficient can be presetted by supplier, as [0.1,1].

For example: just be delithted with the content characteristic of (fancy grade height) originally for a user, the availability coefficient value is 0.9 or 1 to take advantage of described interest-degree; Originally the content characteristic of just disliking for a user (fancy grade is low), the availability coefficient value is that less coefficient such as 0.1 is taken advantage of described interest-degree, reducing the influence of interest-degree, because probably be that the user accompanies friend to watch described content characteristic or other former thereby watched described content characteristic to fancy grade.

Certainly, if the user is higher to the fancy grade of a certain content feature in the files on each of customers, also can increase the influence of described interest-degree to files on each of customers, promptly the magnitude range of described adjustment coefficient is not limited to above-mentioned [0.1,1], also can be greater than 1.

This files on each of customers correcting device 106 is used for according to adjusted interest-degree at the content characteristic that this program presetted, and revises the fancy grade and the weight of content characteristic corresponding in the described files on each of customers, thereby realizes dynamically updating more accurately files on each of customers.

This system also comprises a performance receiving apparatus 101, a screening plant 102 and a files on each of customers management devices 107.

This performance receiving apparatus 101 is used for program receiving information and with the corresponding electronic program guide of digital television of program (EPG) etc.

This screening plant 102 is used for filtering out the program that the user may like according to programme information that is received and files on each of customers, is listed in the described recommendation information tabulation with the programme information that the user may be liked.

This files on each of customers management devices 107 is used for the leading subscriber archives.Described files on each of customers comprises fancy grade and the weight of a user to a plurality of content characteristics usually.

Fig. 2 is schematic flow sheet that upgrades the files on each of customers method according to an embodiment of the invention.

The first step is set up a files on each of customers, and this files on each of customers comprises fancy grade and the weight (step S210) of a user to a plurality of content characteristics.

Wherein said content characteristic is meant each actor names of comprising in the program (Fan Bingbing, Pueraria lobota excellent), program category (literary film, romance movie, horror film etc.), director (Zhang Yimou, Feng Xiaogang etc.) etc., these content characteristics can be from any information sources such as broadcasting, TV or internet, the most representatively sends to the user by electronic program guide of digital television (EPG) with program exactly.

Content characteristic in the files on each of customers can be one, as has only certain performer.Certainly, the content characteristic in the files on each of customers also can have a plurality of, and at this moment, corresponding recommendation results will be more accurate.

Described fancy grade is meant the sensation of user to each content characteristic, can preset a number range by supplier, as [50 ,+50].

Described weight is meant the user when selecting program, various dissimilar content characteristics, as, performer, director, program category etc. are to the influence of its selection result.What also we can say the user when the program of selecting oneself to like, be standard normally with, promptly still directs according to performer, program category and wait the program of selecting it to like.Wherein all performers' weight can be the same, and the weight of all program categories can be the same, and all directors' weight also can be the same.Described weight also can preset a number range by supplier, as [0,100].

Weight in the described files on each of customers and fancy grade, can be described as the historical record that the user watches program, when the user watches program, other history information can be arranged also, watch the number of times of the program that comprises a certain content feature and delete number of times etc. as the user.

Described files on each of customers can be filled in and carried out initialization by user oneself.Certainly, this is not unique mode, can obtain files on each of customers by other approach, as by production firm described commending system being carried out the files on each of customers initialization according to user basic information (as sex, the age) yet.

In above-mentioned files on each of customers, comprise a series of content characteristic.Each content characteristic comprises a ternary array (content characteristic term, favorable rating Like-Degree, weights W eight).Like this, this files on each of customers (User Profile, be called for short UP) can be expressed as a ternary array vector (w), if m different content characteristic arranged in files on each of customers, it can be represented with following this vector array for t, ld:

UP＝((t ₁，ld ₁，w ₁)，(t ₂，ld ₂，w ₂)，…(t _i，ld _i，w _i)…，t _m，ld _m，w _m))(1)

Here t _iBe a content characteristic, i is content characteristic t _iSequence number, ld _iBe for this content characteristic t _iFavorable rating, w _iBe content characteristic t _iWeight.

Supposing a files on each of customers in the current historical record, is such:

Program category: weight=90

Film like-degree=50

Opera like-degree=30

News like-degree=-20, wherein negative number representation is disliked degree,

Then, the user interest ternary array of content characteristic is (film, 50,90) in the above-mentioned program category;

Performer: weight=80

First like-degree=45

Second like-degree=40

Third like-degree=-12.5;

Then, the user is (the third ,-12.5,80) to third user interest ternary array among the above-mentioned performer;

The user is (first, 45,80) to the user interest ternary array of first among the above-mentioned performer

……

Second step, monitor a user behavior, the behavior is at a program (step S220) in broadcast.Described user's behavior comprises that the user watches the time span of the program in the described broadcast, and this program comprises one or more certain content features that preset, and the user watches the number of times of the program that comprises described certain content feature and deletes number of times etc.Described program in broadcast can be that the user elects the program of watching from the recommendation information tabulation.

For example, described program in broadcast is film A, and this film A is a content characteristic that presets, and also comprises other one or more content characteristics in this program, as actor A and performer's third gradegrade C.These content characteristics can be presetted also and can be sent to the user with program by electronic program guide of digital television (EPG) by program provider.

The 3rd step according to this user behavior, obtained the interest-degree of this user to this program, and this interest-degree is at content characteristic (step S230) that this program presetted.Normally play according to the time span of watching described program in the described user behavior, described program predetermination total time length and predetermined threshold value obtain.

Described interest-degree can be expressed as:

WD wherein _iThe time span that the expression user watches described program, θ is a predetermined threshold value, RD _iThe total time length of representing the predetermined broadcast of this specific program.Wherein this preset value can be presetted by supplier, for example, if RD _iBe 2 hours, can establish θ is 0.5 hour, if WD _iLess than 0.5 hour, then putting interest-degree was 0.

For example described program-----film A, establishing its predetermined broadcast total time length is 2 hours, and the θ value is 0.5 hour, and the user has watched 1.5 hours.By formula

Can draw the user is 5 to the interest-degree of this program.We can say that also the user all is 5 to the interest of each content characteristic in this program, promptly the user is 5 to the interest-degree of film A, actor A and performer's third gradegrade C.

The 4th goes on foot, and obtains the fancy grade (step S240) of corresponding content feature in the described files on each of customers.The content characteristic of wherein said correspondence is corresponding with content characteristic in the described program, in this corresponding content characteristic files on each of customers a fancy grade has been arranged.

Certainly, also can obtain the weight of corresponding content feature in the described files on each of customers, this corresponding content characteristic has had a corresponding weights in files on each of customers.

Some content characteristic in the program might not have corresponding content characteristic yet in files on each of customers, the fancy grade of then establishing described some content characteristic is 0, weight is then the same with corresponding type, be the weight that the performer just uses performer in the files on each of customers, program category is just used the weight of program category in the files on each of customers.

For example, actor A in described each content characteristic of film A and performer third are corresponding with content characteristic actor A and performer third in the above-mentioned files on each of customers respectively.The fancy grade of the actor A in above-mentioned files on each of customers is 45, and weight is 80; The fancy grade of performer third in above-mentioned files on each of customers is-12.5, and weight is identical and be 80.

In the 5th step,, this interest-degree is adjusted (step S250) according to the fancy grade of content characteristic corresponding in the described files on each of customers.

Can come above-mentioned interest-degree is adjusted by a coefficient, this coefficient can equal 1 or less than 1 positive decimal, its magnitude range can be presetted by supplier, as [0.1,1] for one; Also can be by dynamically obtaining in conjunction with user preferences degree and other history information, for example, the number of times that can watch described content characteristic in conjunction with fancy grade in the files on each of customers and user is an input variable with the ratio of deletion number of times, utilize mode dynamically to draw above-mentioned coefficient (detailed process please be joined the introduction of aftermentioned Fig. 3) with fuzzy reasoning.

Be example still with above-mentioned film A, for content characteristic actor A wherein, fancy grade in its pairing files on each of customers is 45, as seen the user just to be delithted with this content characteristic actor A originally very interested in the content characteristic actor A in other words, at this moment, an available bigger coefficient is adjusted described interest-degree 5, is 0.9 as this coefficient, and the user becomes 4.5 to the interest-degree of content characteristic actor A after adjusting.The adjustment degree is very little, has reduced the influence of described interest-degree to files on each of customers on the not half.

And for wherein content characteristic performer third, fancy grade in its pairing files on each of customers is-12.5, as seen the user be not originally be delithted with this content characteristic performer third in other words the user content characteristic performer third is lost interest in, at this moment, an available less coefficient is adjusted described interest-degree 5, as this coefficient is 0.3, and the user becomes 1.5 to content characteristic performer third interest-degree after adjusting.The adjustment degree is bigger, has reduced the influence of described interest-degree to files on each of customers to a great extent.

Certainly, if the user is higher to the fancy grade of a certain content feature in the files on each of customers, also can increase the influence of described interest-degree to files on each of customers, promptly the magnitude range of described adjustment coefficient is not limited to above-mentioned [0.1,1], also can be greater than 1.

Above-mentioned adjusted interest-degree will be used to revise corresponding fancy grade and weight, so can adjust coefficient with a fancy grade when adjustment is used for revising the interest-degree of fancy grade; When adjustment is used for revising the interest-degree of weight, adjust coefficient with a weight, these two coefficients are interrelated, influence such as weight adjustment coefficient is subjected to fancy grade to adjust coefficient, and both are proportional etc.Certainly, also can adjust weight and fancy grade simultaneously with the same adjustment coefficient.

The fancy grade of content characteristic can be different in the pairing files on each of customers of described each content characteristic, make the pairing fancy grade of each content characteristic adjust coefficient or weight adjustment coefficient possibility difference, thereby cause the pairing adjusted interest-degree possibility of each content characteristic in the described program different.As, the adjusted interest-degree of described content characteristic actor A is just different with adjusted interest-degree of the content characteristic performer third.

In the 6th step, according to adjusted interest-degree, revise the fancy grade and the weight (step S260) of content characteristic corresponding in the described files on each of customers, thereby realize more dynamically revising files on each of customers at the content characteristic that this program presetted.

The fancy grade and the weight of corresponding content characteristic in the described files on each of customers of above-mentioned correction, can be undertaken by following formula:

${\mathrm{Weight}}_{\mathrm{ti}}^{\′}={\mathrm{Weight}}_{\mathrm{ti}}+{\mathrm{\α}}_t\·\frac{(W{D}_i-\mathrm{\θ})}{R{D}_i}...\left(2\right)$

${\mathrm{Like}\_\mathrm{degree}}_i^{\′}={\mathrm{Like}\_\mathrm{degree}}_i+{\mathrm{\β}}_i\·\frac{({\mathrm{WD}}_i-\mathrm{\θ})}{R{D}_i}...\left(3\right)$

Here, t (Term) represents content characteristic, and i represents a sequence number of content characteristic, i.e. content characteristic i, Weight _TiThe initial weight of expression content characteristic i, Like_degree _iThe expression user is to the initial fancy grade of content characteristic i.Weight ' _TiThe weight of content characteristic i after expression changes, Like_degree ' _iUser preferences degree after expression changes.WD _iBe the actual time span of watching the described program that comprises content characteristic i of described user, RD _iBe the predetermined broadcast total time length of described program itself, so the described predetermined threshold value of θ is θ.

α _tAnd β _iBe respectively weight adjustment factor and fancy grade adjustment factor, wherein β _iWith α _tBecome interrelated, as proportional relation or other relation etc.α _tAnd β _iBe used for adjusting the interest-degree of the affiliated type weight of content characteristic i respectively

And the user is to the interest-degree in the content characteristic i fancy grade

α _tAnd β _iBe generally used for delaying the variation of weight and favorable rating, they are less than or equal to 1 (also can greater than 1), and because the weight of user preferences is relatively stable, so α _t≤ β _i

In calculating process,

If Weight ' _iGreater than 100, make Weight ' _i=100;

If Weight ' _iLess than 0, make Weight ' _i=0;

If Like_degree ' _iGreater than 50, make Like_degree ' _i=50;

If Like_degree ' _iLess than-50, make Like_degree ' _i=-50.

Be that example is illustrated still below with above-mentioned film A:

The user can be expressed as for the fancy grade of actor A and the correction of weight:

Suppose α _t=0.1 β _iβ wherein _iStill aforesaid 0.9, α then _tBe 0.09; The i here represents the content characteristic actor A;

${\mathrm{Like}\_\mathrm{degree}}_i^{\′}={\mathrm{Like}\_\mathrm{degree}}_i+{\mathrm{\β}}_i\·\frac{({\mathrm{WD}}_i-\mathrm{\θ})}{R{D}_i}=45+0.9*5=49.5;$

${\mathrm{Weight}}_{\mathrm{ti}}^{\′}={\mathrm{Weight}}_{\mathrm{ti}}+{\mathrm{\α}}_t\·\frac{(W{D}_i-\mathrm{\θ})}{R{D}_i}=80+0.09*5=80.45.$

The user can be expressed as for performer third the fancy grade and the correction of weight:

α _t=0.3 β _i, β wherein _iStill aforesaid 0.3, α then _tBe 0.09; The i here represents content characteristic performer third;

${\mathrm{Like}\_\mathrm{degree}}_i^{\′}={\mathrm{Like}\_\mathrm{degree}}_i+{\mathrm{\β}}_i\·\frac{({\mathrm{WD}}_i-\mathrm{\θ})}{R{D}_i}=-12.5+0.3*5=-11;$

${\mathrm{Weight}}_{\mathrm{ti}}^{\′}={\mathrm{Weight}}_{\mathrm{ti}}+{\mathrm{\α}}_t\·\frac{(W{D}_i-\mathrm{\θ})}{R{D}_i}=80+0.09*5=80.45.$

All belong to performer's type for actor A and performer third,, be 80.45 so revised weight also is the same.

The weight of same kind (as the performer) normally, and weight depends on α _t, that is depend on adjusted interest-degree.Therefore, only calculate once just passable for the weight of same kind.

The method of the described renewal files on each of customers of present embodiment, after the total time length ratio that utilizes the user to watch the time span of a specific program and this program predetermination to play obtains the interest-degree of user to this program, again user in this interest-degree and the files on each of customers is compared the fancy grade of each content characteristic in this program or other historical record (as the number of times of the specific program that comprises one or more content characteristics as described in watching or deletion number of times etc.), and come described interest-degree is adjusted, thereby obtain user's interest-degree more accurately according to comparative result.

For example, if or the number of times watched not high to the fancy grade of corresponding contents feature seldom in the original files on each of customers, then reduce of the influence of described interest-degree to the fancy grade of described corresponding contents feature; If originally in the files on each of customers to the fancy grade height of corresponding contents feature, then do not reduce of the influence of described interest-degree to the fancy grade of described corresponding contents feature.

Therefore, use the described method of present embodiment to revise files on each of customers, reduced under some special scenes, careless or change platform and watched certain program or accompanied friend to watch certain program etc. as the user, go to revise the possibility of files on each of customers by normal condition, upgrade files on each of customers thereby change according to user interest more accurately.

Fig. 3 is another flow chart that upgrades files on each of customers according to an embodiment of the invention.

The first step with the fancy grade e2 of corresponding contents feature in the files on each of customers and watch the ratio e1 of the number of times that comprises described content characteristic and deletion number of times, is input variable, adjusts factor alpha with weight _tComponent α _IjBe output variable, set up the fuzzy reasoning transformation relation (step S310) of input more than and single output.That is, with fancy grade and other history information of corresponding contents feature in the files on each of customers, present embodiment is meant watches the number of times and deletion number of times that comprises described content characteristic, is used as the input variable of fuzzy reasoning, to obtain output variable α _Ij

Wherein said e1=Pf _i(+)/Pf _i(-), e2=Like-degree _i

Pf _i(+)/Pf _i(-) comes from watching the program that comprises described content characteristic and deletion to comprise the statistics of the program number of times of described content characteristic, concrete condition as shown in Table 2:

In the table, Nf _Gi(+) or Nf _Ai(+) person represents content feature G _i(the content characteristic i of relevant program category) or A _jThe program of (relevant performer content characteristic j) watch number of times, comprise current once record, Nf _Gi(-) or Nf _Aj(-) expression is to content feature G _iOr A _jThe deletion number of times of program, comprise current once record, the user whenever watches content feature G one time _iOr A _iProgram, Nf _Gi(+) or Nf _Aj(+) just increases by 1, content characteristic G of the every deletion of user _iOr A _j, Nf _Gi(-) or Nf _Aj(-) just increases by 1.

Form 2

Classification		Watch number of times	The deletion number of times
				Program category	Content characteristic 1	Nf G1(+)	Nf G1(-)
.....	......	......
			Content characteristic i		Nf Gi(+)	Nf Gi(-)
......	......	......
			Content characteristic m		Nf Gm(+)	Nf Gm(-)
The performer	Content characteristic	1					Nf A1(+)	Nf A1(-)
			......	......	......
	Content characteristic j	Nf Aj(+)				Nf Aj(-)
			......	......	......
	Content characteristic l	Nf Al(+)				Nf Al(-)

By making Pf _i(+)=Nf _i(+)/Nf (+) and Pf _i(-)=Nf _i(-)/Nf (-) obtains ratio Pf _i(+)/Pf _i(-).

The number of times of the program that comprises all the elements feature is watched in Nf (+) expression, and Nf (-) expression deletion comprises the number of times of all the elements feature program.That is Nf (+)=∑ Nf, _i(+), Nf (-)=∑ Nf _i(-).

In second step,, adjust the fuzzy value (step S320) of coefficient component to draw described weight by fuzzy reasoning.

In the process of above-mentioned fuzzy reasoning, suppose program is divided into n time slice, then this α _JiExpression content characteristic i component of pairing weight adjustment coefficient during j time slice in a program.

By the variable relation of above-mentioned many inputs and single output, utilize fuzzy reasoning to draw weight and adjust coefficient component α _JiFuzzy value.In the present embodiment, try to achieve α by obfuscation e1 and e2 _JiFuzzy value.

Concrete reasoning process is illustrated below in conjunction with Fig. 4, Fig. 5 and Fig. 6.Wherein Fig. 4 represents the fuzzy curve chart of input variable e1, and Fig. 5 represents the fuzzy curve chart of input variable e2, and Fig. 6 represents according to input variable e1 and e2, carries out fuzzy reasoning and the output variable α that draws _JiFuzzy curve chart.The degree of consistency of the new and old interest of user (variation of user interest) has reflected the degree that the user interest degree need be adjusted, if new and old interest consistency is low, then needs to do a large amount of adjustment, so α _JiLess, otherwise α _JiBigger.Thus, it is as follows to draw concrete fuzzy inference rule:

If I. e1 is " greatly ", and e2 is " liking ", α so _JiBe " greatly ";

If II. e1 is " greatly ", and e2 is " generally ", α so _JiBe " bigger ";

If III. e1 is " greatly ", and e2 is " disliking ", α so _JiBe " medium ";

If IV. e1 is " medium ", and e2 is " liking ", α so _JiBe " bigger ";

If V. e1 is " medium ", and e2 also is " generally ", α so _JiBe " medium ";

If VI. e1 is " medium ", and e2 is " disliking ", α so _JiBe " less ";

If VII. e1 is " little ", and e2 is " liking ", α so _JiBe " medium ";

If VIII. e1 is " little ", and e2 is " generally ", α so _JiBe " less ";

If IX. e1 is " little ", and e2 is " disliking ", α so _JiBe " little ";

Wherein the μ value representation e1 among Fig. 4 and Fig. 5, the degree of membership of e2, the degree of membership μ among Fig. 6 be according to e1 among Fig. 4 and Fig. 5 and e2 degree of membership obtain.

In the 3rd step, obtain the clearly value (step S330) that weight is adjusted the component of coefficient.That is, above-mentioned weight is adjusted coefficient component α _JiFuzzy value, obtain behind the de-fuzzy adjusting with heavy coefficient component α _JiClearly value.

Be convenient to understand for final result, the result of fuzzy reasoning must convert clear amount to.Modal deblurring algorithm is area gravity model appoach and maximum average value method.The former synthesizes the rule of all excitation outputs as a result of, is applicable to level and smooth control, is process control method commonly used.

Present embodiment adopts area center of gravity deblurring algorithm, and this algorithm as shown in Equation (4).

${\mathrm{\α}}_{\mathrm{ji}}=\underset{l=1}{\overset{p}{\mathrm{\Σ}}}\mathrm{\μ}\left[l\right],y_l/\underset{l=1}{\overset{p}{\mathrm{\Σ}}}\mathrm{\μ}\left[l\right]...\left(4\right)$

Here, μ [l]: the height of output area is inferred in expression from l rule;

Yl: the abscissa of the center of gravity of output area is inferred in expression from l rule;

P: the quantity of rule of inference is satisfied in expression.

Utilize above-mentioned formula, can obtain α _JiClearly value, detailed process can be with reference to No. the 200310123354.7th, Chinese patent application.

In the 4th step, obtain weight and adjust coefficient (step S340).This step comprises following two processes:

A, weighting recanalization coefficient component α _JiClearly the value mean value, to draw α _Tj), α wherein _TjRepresent a corresponding α of all the elements feature that belongs to t type (as performer, director, program category etc.) _JiMean value, promptly the weight of t type when each time slice adjusted factor alpha _TjThis step can be realized by following formula:

${\mathrm{\&alpha;}}_{\mathrm{tj}}=\frac{{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="tj"\; filenames="A20058003938700251.png,A20058003938700252.png"\; state="\{\{state\}\}">tj</figure-callout>}1}+\&CenterDot;\&CenterDot;\&CenterDot;{\mathrm{\&alpha;}}_{\mathrm{tji}}+\&CenterDot;\&CenterDot;\&CenterDot;+{\mathrm{\&alpha;}}_{\mathrm{tjm}}}{m}...\left(5\right)$

Wherein, above-mentioned m represents that m content characteristic arranged in the t category feature.

B, based on the mean value that is obtained, obtain weight and adjust coefficient.The weight of this H category information is adjusted number α _tCan obtain by following formula:

${\mathrm{\&alpha;}}_t=\frac{{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A20058003938700251.png,A20058003938700252.png"\; state="\{\{state\}\}">t</figure-callout>}1}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{j}{\mathrm{\&alpha;}}_{\mathrm{ij}}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{n}{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="tn"\; filenames="A20058003938700251.png,A20058003938700252.png"\; state="\{\{state\}\}">tn</figure-callout>}}}{1+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{j}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{n}}...\left(6\right)$

Wherein, n express time number of fragments.

In the 5th step, obtain fancy grade and adjust coefficient (step S350).

Clearly value based on the weight adjustment coefficient component that is obtained draws fancy grade and adjusts coefficient.This fancy grade is adjusted factor beta _iCan obtain by following formula:

${\mathrm{\&beta;}}_i=\frac{{\mathrm{\&alpha;}}_{1i}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{j}{\mathrm{\&alpha;}}_{\mathrm{ji}}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{n}{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="1"\; label="ni"\; filenames="A20058003938700251.png,A20058003938700252.png"\; state="\{\{state\}\}">ni</figure-callout>}}}{1+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{j}+\&CenterDot;\&CenterDot;\&CenterDot;+\frac{1}{n}}...\left(7\right)$

Wherein, n express time number of fragments, i represents content characteristic i.

In the 6th step, adjust factor alpha according to described weight _tAnd fancy grade is adjusted factor beta _i, this interest-degree is adjusted (step S360) accordingly.

In the 7th step,, revise the fancy grade and the weight (step S370) of content characteristic corresponding in the described files on each of customers according to adjusted interest-degree at the content characteristic that this program presetted.

Though through the present invention is described in conjunction with specific embodiments, for the skilled personage in present technique field, will be conspicuous according to manyly substituting of making of narration above, modifications and variations.Therefore, when such substituting, in the spirit and scope that modifications and variations fall into attached claim the time, should being included among the present invention.

Claims (14)

1. method of upgrading files on each of customers, this files on each of customers comprises the fancy grade of a user at least one content characteristic, comprises step:

(a). monitor this user's behavior, the behavior is at a program in broadcast;

(b). according to this user behavior, obtain the interest-degree of this user to this program, this interest-degree is at content characteristic that this program presetted;

(c). according to the fancy grade of content characteristic corresponding in the described files on each of customers, this interest-degree is correspondingly adjusted; With

(d). according to adjusted interest-degree, revise the fancy grade of content characteristic corresponding in the described files on each of customers at the content characteristic that this program presetted.

2. the method for claim 1, wherein said files on each of customers comprises the weight of user to described at least one content characteristic, described method also comprises step: according to adjusted interest-degree at the content characteristic that this program presetted, revise the weight of the content characteristic of described files on each of customers correspondence.

3. the method for claim 1 wherein also comprises step in the step (c):

(i) obtain in the described files on each of customers fancy grade of corresponding content characteristic,

If (ii) this fancy grade shows that the user loses interest in to this content characteristic, then adjust described interest-degree, to reduce of the influence of this interest-degree to described files on each of customers.

4. method as claimed in claim 3, wherein step comprises step in (ii): adjust described interest-degree by taking advantage of a coefficient less than 1, to reduce the influence of described user interest degree to described files on each of customers.

5. the method for claim 1 wherein also comprises step in the step (c):

Obtain the fancy grade of content characteristic corresponding in the described files on each of customers,

If it is interested in this content characteristic that described fancy grade shows the user, then adjust described interest-degree, to increase of the influence of this interest-degree to described files on each of customers.

6. the method for claim 1 wherein also comprises step in the step (c):

Obtain the fancy grade of content characteristic corresponding in the described files on each of customers,

According to this fancy grade, utilize the mode of fuzzy reasoning to adjust described interest-degree.

7. the method for claim 1, wherein said user behavior comprises that the user watches the time span of described program.

8. device that upgrades files on each of customers, this files on each of customers comprises the fancy grade of a user at least one content characteristic, and this device comprises:

A user interaction means is used to monitor this user's behavior, and the behavior is at a program in broadcast;

An interests change analytical equipment is used for according to this user behavior, obtains the interest-degree of this user to this program, and this interest-degree is at content characteristic that this program presetted;

An interests change adjusting device is used for the fancy grade according to the corresponding content characteristic of described files on each of customers, and this interest-degree is correspondingly adjusted; With

A files on each of customers correcting device is used for according to adjusted interest-degree at the content characteristic that this program presetted, and revises the fancy grade of content characteristic corresponding in the described files on each of customers.

9. device as claimed in claim 8, wherein said files on each of customers comprises the weight of user to described a plurality of content characteristics, described files on each of customers correcting device also is used for revising the weight of the content characteristic of described files on each of customers correspondence according to adjusted interest-degree at the content characteristic that this program presetted.

10. device as claimed in claim 8, wherein the interests change adjusting device also is used for obtaining the fancy grade of content characteristic corresponding in the described files on each of customers, if this fancy grade shows the user this content characteristic is lost interest in, then adjust described interest-degree, to reduce of the influence of this interest-degree to described files on each of customers.

11. device as claimed in claim 8, wherein said interests change adjusting device also is used for obtaining the fancy grade of the corresponding content characteristic of described files on each of customers, if it is interested in this content characteristic that described fancy grade shows the user, then adjust described interest-degree, to increase of the influence of this interest-degree to described files on each of customers.

12. device as claimed in claim 8, wherein said interests change adjusting device also are used for obtaining the fancy grade of content characteristic corresponding in the described files on each of customers, according to this fancy grade, utilize the mode of fuzzy reasoning to adjust described interest-degree.

13. device as claimed in claim 8, wherein said user behavior comprise that the user watches the time span of described program.

14. an information recommendation system comprises:

A performance receiving apparatus is used for program receiving information;

A files on each of customers management devices is used for the stored user archives, and this files on each of customers comprises the fancy grade of a user at least one content characteristic;

A screening plant is used for selecting the programme information that the user may like according to described files on each of customers from described programme information, gives the user to recommend this programme information of selecting;

A user interaction means is used to monitor this user's behavior, and the behavior is at institute's recommend programs information;

An interests change analytical equipment is used for according to this user behavior, obtains the interest-degree of this user to this program, and this interest-degree is at content characteristic that this program presetted;

An interests change adjusting device is used for the fancy grade according to the corresponding content characteristic of described files on each of customers, and this interest-degree is correspondingly adjusted; With

A files on each of customers correcting device is used for according to adjusted interest-degree at the content characteristic that this program presetted, and revises the fancy grade of content characteristic corresponding in the described files on each of customers.

Images