CN104572825A - Method and device for recommending information - Google Patents
Method and device for recommending information Download PDFInfo
- Publication number
- CN104572825A CN104572825A CN201410738395.5A CN201410738395A CN104572825A CN 104572825 A CN104572825 A CN 104572825A CN 201410738395 A CN201410738395 A CN 201410738395A CN 104572825 A CN104572825 A CN 104572825A
- Authority
- CN
- China
- Prior art keywords
- recommended
- content
- similarity
- information
- query information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Abstract
The invention discloses a method and a device for recommending information. The method for recommending the information comprises the following steps of obtaining query information and corresponding to-be-recommended candidate content; obtaining the parameter information of a learning network corresponding to the query information and the to-be-recommended candidate content; according to the parameter information, calculating the similarity of the query information and the to-be-recommended candidate content in a preset space, and according to the similarity, sorting the to-be-recommended content from the to-be-recommended candidate content, so as to display the to-be-recommended content to a user. The method disclosed by the embodiment has the advantage that by obtaining the query information and the to-be-recommended candidate content corresponding the query information, the parameter information of the learning network corresponding to the query information and the to-be-recommended candidate content is obtained; according to the parameter information, the similarity of the query information and the to-be-recommended candidate content in the preset space is calculated; according to the similarity, the to-be-recommended content can be sorted from the to-be-recommended candidate content, so the recommending content capable of stimulating a user demand can be recommended to the user, and the problem of homogenization is solved.
Description
Technical field
The present invention relates to Internet technical field, particularly relate to a kind of recommend method and device of information.
Background technology
Along with the high speed development of internet, people by search engine retrieving to required resource, thus can meet the demand of self.In order to the demand of user can be excited, a very important function has been become in current internet product to user's content recommendation, such as, search engine while Search Results is shown to user in the left side of search results pages, can recommend more content on the right side of search results pages to user; Or when user browses one section of news on news website time, news website also all can recommend more content by the mode of related news, recommended article to user.
At present, method to user's content recommendation mainly contains two kinds: first method, direct calculating main contents (search word, headline etc. as input) and the similarity of entry to be recommended in content recommendation storehouse, to the content that user recommends similarity higher; Second method, based on collaborative filtering, namely to the content that active user recommends other users similar to the interest, browsing histories etc. of active user browsed.
But, the shortcoming of first method is, the main contents homogeneity that the content of recommending to user and user browse is serious (such as, if the content that user's browsed " apple 6 " is relevant, related news that the content of then recommending is " apple 6 "), be unfavorable for the more demands exciting user; The shortcoming of second method is, at the beginning of product is reached the standard grade, does not have abundant user, cannot form the effect of collaborative filtering, therefore cannot realize the recommendation of collaborative filtering.
Summary of the invention
The present invention is intended to solve one of technical matters in correlation technique at least to a certain extent.For this reason, one object of the present invention is the recommend method proposing a kind of information, and the method can realize the content recommendation recommending out to excite user's request to user, solves the problem of homogeneity.
Second object of the present invention is the recommendation apparatus proposing a kind of information.
For reaching above-mentioned purpose, embodiment proposes a kind of recommend method of information according to a first aspect of the present invention, comprising: the alternating content to be recommended obtaining Query Information and correspondence thereof; Obtain the parameter information of described Query Information and learning network corresponding to described alternating content to be recommended; And calculate described Query Information and the similarity of described alternating content to be recommended in pre-set space according to described parameter information, and from alternating content to be recommended, filter out content to be recommended according to described similarity, for representing described content to be recommended to user.
The recommend method of the information of the embodiment of the present invention, by obtaining the alternating content to be recommended of Query Information and correspondence thereof, obtain the parameter information of Query Information and learning network corresponding to alternating content to be recommended, and calculate Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and from alternating content to be recommended, filter out content to be recommended according to similarity, content to be recommended is represented for user, the content recommendation recommending out to excite user's request to user can be realized, solve the problem of homogeneity.
For reaching above-mentioned purpose, embodiment proposes a kind of recommendation apparatus of information according to a second aspect of the present invention, comprising: first obtains module, for obtaining the alternating content to be recommended of Query Information and correspondence thereof; Second obtains module, for obtaining the parameter information of described Query Information and learning network corresponding to described alternating content to be recommended; And screening module, for calculating described Query Information and the similarity of described alternating content to be recommended in pre-set space according to described parameter information, and from alternating content to be recommended, filter out content to be recommended according to described similarity, for representing described content to be recommended to user.
The recommendation apparatus of the information of the embodiment of the present invention, by obtaining the alternating content to be recommended of Query Information and correspondence thereof, obtain the parameter information of Query Information and learning network corresponding to alternating content to be recommended, and calculate Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and from alternating content to be recommended, filter out content to be recommended according to similarity, content to be recommended is represented for user, the content recommendation recommending out to excite user's request to user can be realized, solve the problem of homogeneity.
Additional aspect of the present invention and advantage will part provide in the following description, and part will become obvious from the following description, or be recognized by practice of the present invention.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the recommend method of information according to an embodiment of the invention.
Fig. 2 is learning network schematic diagram according to an embodiment of the invention.
Fig. 3 is sigmoid function curve effect schematic diagram according to an embodiment of the invention.
Fig. 4 is the process flow diagram of the recommend method of information according to the present invention's specific embodiment.
Fig. 5 is the commending system schematic diagram of the information according to the present invention's specific embodiment.
Fig. 6 is the structural representation of the recommendation apparatus of information according to an embodiment of the invention.
Embodiment
Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Be exemplary below by the embodiment be described with reference to the drawings, be intended to for explaining the present invention, and can not limitation of the present invention be interpreted as.
Below with reference to the accompanying drawings recommend method and the device of the information of the embodiment of the present invention are described.
Fig. 1 is the process flow diagram of the recommend method of information according to an embodiment of the invention.
As shown in Figure 1, the recommend method of this information comprises:
S101, obtains the alternating content to be recommended of Query Information and correspondence thereof.
In an embodiment of the present invention, the alternating content to be recommended of Query Information and correspondence thereof can be obtained.Suppose that Query Information is Q, the alternating content to be recommended of its correspondence can be C, and wherein, C is C1, C2, C3 ... the set of Cn.
S102, obtains the parameter information of Query Information and learning network corresponding to alternating content to be recommended.
Particularly, corresponding learning network can be built according to Query Information and alternating content to be recommended, then obtain the training sample of learning network, use training sample to be got parms information by preset algorithm such as back-propagation algorithm, stochastic gradient descent methods.
Particularly, as shown in Figure 2, suppose that learning network is three layers, then the step of get parms information W1, W2 can be divided into following three steps.
The first step: obtain training sample <T1, Image (T1) >, <T2, Image (T2) > ... <Tn, Image (Tn) >, wherein, T is text, and Image (T) is corresponding picture.
Second step: obtain <<Ta1 at random or according to predetermined filtering strategy, Image (Ta1) >, <Tb1, Image (Tb1) >>,
<<Ta2,Image(Ta2)>,<Tb2,Image(Tb2)>>...
<<Tam,Image(Tam)>,<Tbm,Image(Tbm)>>，
Calculate picture analogies degree <Ta1, Tb1, ImageSim (Image (Ta1), Image (Tb1) >,
<Ta2,Tb2,ImageSim(Image(Ta2),Image(Tb2)>…
<Tam,Tbm,ImageSim(Image(Tam),Image(Tbm)>。Wherein, ImageSim (Image (Ta), Image (Tb)) represents the similarity between Image (Ta) and Image (Tb).
3rd step: by gradient (the parameter information W1 of back-propagation algorithm (Back Propagation) calculation training sample, the partial derivative of W2), by stochastic gradient descent method (SGD, Stochastic Gradient Descent) iteration upgrades the parameter information W1 that crosses of random initializtion, W2.Particularly, the method that iteration upgrades is W1, W2 deducts default learning rate, be multiplied by the gradient calculated again, make picture analogies degree ImageSim (Image (Ta), Image (Tb)) and similarity Sim (Ta, Tb) between square error minimize.
Herein, adopt matching picture similarity ImageSim (Image (Ta), Image (Tb)) to calculate similarity Sim (Ta, Tb), realize the similarity of heterogeneousization.The factor affecting similarity heterogeneousization can have multiple, and one of them factor is that the visual primitive characters such as RGB (RGB, Red Green Blue), ensure that the accuracy of picture analogies degree result of calculation because contain gray scale in picture.And similarity Sim (Ta, Tb) is abstract feature, the primitive character comparing picture is sparse, cannot ensure the accuracy of result of calculation.Therefore, the similarity in pre-set space according to Query Information and alternating content to be recommended, can recommend out not to be according to similarity Sim (Ta, Tb) content recommendation directly calculated, but the content recommendation that recommendation is calculated according to picture analogies degree ImageSim (Image (Ta), Image (Tb)).According to the parameter information W1 that matching picture Similarity Measure goes out, W2, text can be transformed into pre-set space, thus carry out content recommendation according to the similarity calculated in pre-set space, improve the diversity of content recommendation.
It should be noted that, the pre-set space in the present embodiment can be imagination space.Imagination space is when seeing a character string based on user, forms the picture of an imagination in the brain, represents by the form of picture the meaning that this character string will be expressed.Therefore, the mathematical notation that character string transfers in higher dimensional space, thus calculate the similarity of character string at higher dimensional space, higher dimensional space is herein imagination space.
S103, calculates Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and from alternating content to be recommended, filters out content to be recommended according to similarity, represent content to be recommended for user.
In an embodiment of the present invention, Query Information primary vector corresponding in pre-set space can be calculated according to parameter information, and calculate each alternating content to be recommended secondary vector corresponding in pre-set space according to parameter information, then the similarity between primary vector and each secondary vector is calculated, thus Query Information and the similarity of alternating content to be recommended in pre-set space can be calculated, by the similarity between primary vector and each secondary vector as Query Information and alternating content to be recommended similarity in pre-set space.
Be illustrated with primary vector V (Q), suppose that Query Information Q is for " I loves Tian An-men, Beijing ", learning network is three layers, then ground floor input layer is made up of " I ", " love ", " Beijing " and " Tian An-men " four words, the size of dictionary V is 100000, then be input as the vectorial V1 of 100000 dimensions, the value of vector " I ", " love ", " Beijing " of V1, " Tian An-men " this position corresponding to four words is 1, and the value corresponding to the position of other words in the middle of dictionary V is 0.The second layer is the hidden layer of learning network, and vectorial V2, the Layer2_SIZE of to be a size be Layer2_SIZE are integers being greater than 1, and span is 100 to 10000.The value of each element V2 [i] in vector V2, draws by following formulae discovery:
V2[i]＝sigmoid(W1[i][0]*V1[0]+W1[i][1]*V1[1]+…+W1[i][100000-1]*V1[100000-1])
Wherein, sigmoid function is a function being used for carrying out nonlinear transformation
its function curve can be as shown in Figure 3.Third layer is that the common value of the high dimension vector V3 carrying out Similarity Measure, its size Layer3_Size can be less than Layer2_Size, and span is 50 to 2000.Wherein, the computing method of each element draw by following formulae discovery:
V3[i]＝sigmoid(W2[i][0]*V2[0]+W2[i][1]*V2[1]+…+W2[i][Layer2_SIZE-1]*V2[Layer2_SIZE-1])
The V3 finally calculated is V (Q), i.e. primary vector.
In like manner, the high dimension vector V (C) of alternating content C to be recommended can be calculated, i.e. secondary vector.
Acquisition primary vector be V (Q), secondary vector is after V (C), then by formula
Calculate the similarity between primary vector and each secondary vector.
After calculating Query Information and the similarity of alternating content to be recommended in pre-set space, content to be recommended can be filtered out according to similarity from alternating content to be recommended.
Particularly, can determine that similarity is greater than the secondary vector of predetermined threshold value, and obtain content to be recommended according to the secondary vector determined, also can determine the secondary vector of predetermined quantity according to similarity order from high to low, and obtain content to be recommended according to the secondary vector determined.
The recommend method of the information of the embodiment of the present invention, by obtaining the alternating content to be recommended of Query Information and correspondence thereof, obtain the parameter information of Query Information and learning network corresponding to alternating content to be recommended, and calculate Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and from alternating content to be recommended, filter out content to be recommended according to similarity, content to be recommended is represented for user, the content recommendation recommending out to excite user's request to user can be realized, solve the problem of homogeneity.
Fig. 4 is the process flow diagram of the recommend method of information according to the present invention's specific embodiment.
As shown in Figure 4, the recommend method of this information comprises:
S401, obtains the alternating content to be recommended of Query Information and correspondence thereof.
Particularly, can obtain Query Information Q, the alternating content to be recommended of its correspondence can be C, and wherein, C is C1, C2, C3 ... the set of Cn.
S402, calculates the Query Information high dimension vector corresponding with alternating content to be recommended according to parameter information.
Particularly, according to parameter information W1, W2, the high dimension vector V (Q) of Query Information can be calculated by forward direction algorithm, the high dimension vector V (C1) that alternating content to be recommended is corresponding, V (C2) ... V (Cn).
Be described for the high dimension vector V (Q) of Query Information, suppose that Q is for " I loves Tian An-men, Beijing ", learning network is three layers, then ground floor input layer is made up of " I ", " love ", " Beijing " and " Tian An-men " four words, the size of dictionary V is 100000, then be input as the vectorial V1 of 100000 dimensions, the value of vector " I ", " love ", " Beijing " of V1, " Tian An-men " this position corresponding to four words is 1, and the value corresponding to the position of other words in the middle of dictionary V is 0.The second layer is the hidden layer of learning network, and vectorial V2, the Layer2_SIZE of to be a size be Layer2_SIZE are integers being greater than 1, and span is 100 to 10000.The value of each element V2 [i] in vector V2, draws by following formulae discovery:
V2[i]＝sigmoid(W1[i][0]*V1[0]+W1[i][1]*V1[1]+…+W1[i][100000-1]*V1[100000-1])
Wherein, sigmoid function is a function being used for carrying out nonlinear transformation
its function curve can be as shown in Figure 3.Third layer is that the common value of the high dimension vector V3 carrying out Similarity Measure, its size Layer3_Size can be less than Layer2_Size, and span is 50 to 2000.Wherein, the computing method of each element draw by following formulae discovery:
V3[i]＝sigmoid(W2[i][0]*V2[0]+W2[i][1]*V2[1]+…+W2[i][Layer2_SIZE-1]*V2[Layer2_SIZE-1])
The V3 finally calculated is V (Q).
In like manner, the high dimension vector V (C) that alternating content to be recommended is corresponding can be calculated.
S403, the high dimension vector corresponding with alternating content to be recommended according to Query Information calculates Query Information and the similarity of alternating content to be recommended in pre-set space.
Particularly, can according to formula
Calculate the high dimension vector V (Q) of the Query Information high dimension vector V (C1) corresponding with each alternating content to be recommended, V (C2) ... similarity Sim (the V (Q) of V (Cn), V (C1)), Sim (V (Q), V (C2)) ... Sim (V (Q), V (Cn)).
S404, filters out content to be recommended according to similarity from alternating content to be recommended.
Particularly, can according to similarity from high to low sequentially screen out content to be recommended, also can according to the content to be recommended filtering out similarity and be greater than predetermined threshold value.
The implementation procedure of above-described embodiment, clearly illustrates that by Fig. 5.
The recommend method of the information of the embodiment of the present invention, by obtaining the alternating content to be recommended of Query Information and correspondence thereof, the Query Information high dimension vector corresponding with alternating content to be recommended is calculated according to parameter information, then corresponding with alternating content to be recommended according to Query Information high dimension vector calculates Query Information and the similarity of alternating content to be recommended in pre-set space, and from alternating content to be recommended, filter out content to be recommended according to similarity, the content recommendation recommending out to excite user's request to user can be realized, solve the problem of homogeneity.
In order to realize above-described embodiment, the present invention also proposes a kind of recommendation apparatus of information.
Fig. 6 is the structural representation of the recommendation apparatus of information according to an embodiment of the invention.
As shown in Figure 6, the recommendation apparatus of this information can comprise: first obtains module 110, second obtains module 120 and screening module 130.
Wherein, first module 110 is obtained for obtaining the alternating content to be recommended of Query Information and correspondence thereof.
In an embodiment of the present invention, first the alternating content to be recommended that module 110 can obtain Query Information and correspondence thereof is obtained.Suppose that Query Information is Q, the alternating content to be recommended of its correspondence can be C, and wherein, C is C1, C2, C3 ... the set of Cn.
Second obtains module 120 for obtaining the parameter information of Query Information and learning network corresponding to alternating content to be recommended.
Particularly, second obtains module 120 can build corresponding learning network according to Query Information and alternating content to be recommended, then obtain the training sample of learning network, use training sample to be got parms information by preset algorithm such as back-propagation algorithm, stochastic gradient descent methods.
Particularly, as shown in Figure 2, suppose that learning network is three layers, then the step of get parms information W1, W2 can be divided into following three steps.
The first step: obtain training sample <T1, Image (T1) >, <T2, Image (T2) > ... <Tn, Image (Tn) >, wherein, T is text, and Image (T) is corresponding picture.
Second step: obtain <<Ta1 at random or according to predetermined filtering strategy, Image (Ta1) >, <Tb1, Image (Tb1) >>,
<<Ta2,Image(Ta2)>,<Tb2,Image(Tb2)>>...
<<Tam,Image(Tam)>,<Tbm,Image(Tbm)>>，
Calculate picture analogies degree <Ta1, Tb1, ImageSim (Image (Ta1), Image (Tb1) >,
<Ta2,Tb2,ImageSim(Image(Ta2),Image(Tb2)>…
<Tam,Tbm,ImageSim(Image(Tam),Image(Tbm)>。Wherein, ImageSim (Image (Ta), Image (Tb)) represents the similarity between Image (Ta) and Image (Tb).
3rd step: by gradient (the parameter information W1 of back-propagation algorithm (Back Propagation) calculation training sample, the partial derivative of W2), by stochastic gradient descent method (SGD, Stochastic Gradient Descent), carry out the parameter information W1 that iteration renewal random initializtion is crossed, W2.Particularly, the method that iteration upgrades is W1, W2 deducts default learning rate, be multiplied by the gradient calculated again, make picture analogies degree ImageSim (Image (Ta), Image (Tb)) and similarity Sim (Ta, Tb) between square error minimize.
Herein, adopt matching picture similarity ImageSim (Image (Ta), Image (Tb)) to calculate similarity Sim (Ta, Tb), realize the similarity of heterogeneousization.The factor affecting similarity heterogeneousization can have multiple, and one of them factor is that the visual primitive characters such as RGB (RGB, Red Green Blue), ensure that the accuracy of picture analogies degree result of calculation because contain gray scale in picture.And similarity Sim (Ta, Tb) is abstract feature, the primitive character comparing picture is sparse, cannot ensure the accuracy of result of calculation.Therefore, the similarity in pre-set space according to Query Information and alternating content to be recommended, can recommend out not to be according to similarity Sim (Ta, Tb) content recommendation directly calculated, but the content recommendation that recommendation is calculated according to picture analogies degree ImageSim (Image (Ta), Image (Tb)).According to the parameter information that matching picture Similarity Measure goes out, text can be transformed into pre-set space, thus carry out content recommendation according to the similarity calculated in pre-set space, improve the diversity of content recommendation.
It should be noted that, the pre-set space in the present embodiment can be imagination space.Imagination space is when seeing a character string based on user, forms the picture of an imagination in the brain, represents by the form of picture the meaning that this character string will be expressed.Therefore, the mathematical notation that character string transfers in higher dimensional space, thus calculate the similarity of character string at higher dimensional space, higher dimensional space is herein imagination space.
Screening module 130 for calculating Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and filters out content to be recommended according to similarity from alternating content to be recommended, represents content to be recommended for user.
In an embodiment of the present invention, screening module 130 can calculate Query Information primary vector corresponding in pre-set space according to parameter information, and calculate each alternating content to be recommended secondary vector corresponding in pre-set space according to parameter information, then the similarity between primary vector and each secondary vector is calculated, thus Query Information and the similarity of alternating content to be recommended in pre-set space can be calculated, by the similarity between primary vector and each secondary vector as Query Information and alternating content to be recommended similarity in pre-set space.
Be illustrated with primary vector V (Q), suppose that Query Information Q is for " I loves Tian An-men, Beijing ", learning network is three layers, then ground floor input layer is made up of " I ", " love ", " Beijing " and " Tian An-men " four words, the size of dictionary V is 100000, then be input as the vectorial V1 of 100000 dimensions, the value of vector " I ", " love ", " Beijing " of V1, " Tian An-men " this position corresponding to four words is 1, and the value corresponding to the position of other words in the middle of dictionary V is 0.The second layer is the hidden layer of learning network, and vectorial V2, the Layer2_SIZE of to be a size be Layer2_SIZE are integers being greater than 1, and span is 100 to 10000.The value of each element V2 [i] in vector V2, draws by following formulae discovery:
V2[i]＝sigmoid(W1[i][0]*V1[0]+W1[i][1]*V1[1]+…+W1[i][100000-1]*V1[100000-1])
Wherein, sigmoid function is a function being used for carrying out nonlinear transformation
its function curve can be as shown in Figure 3.Third layer is that the common value of the high dimension vector V3 carrying out Similarity Measure, its size Layer3_Size can be less than Layer2_Size, and span is 50 to 2000.Wherein, the computing method of each element draw by following formulae discovery:
V3[i]＝sigmoid(W2[i][0]*V2[0]+W2[i][1]*V2[1]+…+W2[i][Layer2_SIZE-1]*V2[Layer2_SIZE-1])
The V3 finally calculated is V (Q), i.e. primary vector.
In like manner, the high dimension vector V (C) of alternating content C to be recommended can be calculated, i.e. secondary vector.
Acquisition primary vector be V (Q), secondary vector is after V (C), then by formula
Calculate the similarity between primary vector and each secondary vector.
After calculating Query Information and the similarity of alternating content to be recommended in pre-set space, screening module 130 can filter out content to be recommended according to similarity from alternating content to be recommended.
Particularly, screening module 130 can determine that similarity is greater than the secondary vector of predetermined threshold value, and obtain content to be recommended according to the secondary vector determined, also can determine the secondary vector of predetermined quantity according to similarity order from high to low, and obtain content to be recommended according to the secondary vector determined.
The recommendation apparatus of the information of the embodiment of the present invention, by obtaining the alternating content to be recommended of Query Information and correspondence thereof, obtain the parameter information of Query Information and learning network corresponding to alternating content to be recommended, and calculate Query Information and the similarity of alternating content to be recommended in pre-set space according to parameter information, and from alternating content to be recommended, filter out content to be recommended according to similarity, content to be recommended is represented for user, the content recommendation recommending out to excite user's request to user can be realized, solve the problem of homogeneity.
In the description of this instructions, specific features, structure, material or feature that the description of reference term " embodiment ", " some embodiments ", " example ", " concrete example " or " some examples " etc. means to describe in conjunction with this embodiment or example are contained at least one embodiment of the present invention or example.In this manual, to the schematic representation of above-mentioned term not must for be identical embodiment or example.And the specific features of description, structure, material or feature can combine in one or more embodiment in office or example in an appropriate manner.In addition, when not conflicting, the feature of the different embodiment described in this instructions or example and different embodiment or example can carry out combining and combining by those skilled in the art.
In addition, term " first ", " second " only for describing object, and can not be interpreted as instruction or hint relative importance or imply the quantity indicating indicated technical characteristic.Thus, be limited with " first ", the feature of " second " can express or impliedly comprise at least one this feature.In describing the invention, the implication of " multiple " is at least two, such as two, three etc., unless otherwise expressly limited specifically.
Describe and can be understood in process flow diagram or in this any process otherwise described or method, represent and comprise one or more for realizing the module of the code of the executable instruction of the step of specific logical function or process, fragment or part, and the scope of the preferred embodiment of the present invention comprises other realization, wherein can not according to order that is shown or that discuss, comprise according to involved function by the mode while of basic or by contrary order, carry out n-back test, this should understand by embodiments of the invention person of ordinary skill in the field.
In flow charts represent or in this logic otherwise described and/or step, such as, the sequencing list of the executable instruction for realizing logic function can be considered to, may be embodied in any computer-readable medium, for instruction execution system, device or equipment (as computer based system, comprise the system of processor or other can from instruction execution system, device or equipment instruction fetch and perform the system of instruction) use, or to use in conjunction with these instruction execution systems, device or equipment.With regard to this instructions, " computer-readable medium " can be anyly can to comprise, store, communicate, propagate or transmission procedure for instruction execution system, device or equipment or the device that uses in conjunction with these instruction execution systems, device or equipment.The example more specifically (non-exhaustive list) of computer-readable medium comprises following: the electrical connection section (electronic installation) with one or more wiring, portable computer diskette box (magnetic device), random access memory (RAM), ROM (read-only memory) (ROM), erasablely edit ROM (read-only memory) (EPROM or flash memory), fiber device, and portable optic disk ROM (read-only memory) (CDROM).In addition, computer-readable medium can be even paper or other suitable media that can print described program thereon, because can such as by carrying out optical scanning to paper or other media, then carry out editing, decipher or carry out process with other suitable methods if desired and electronically obtain described program, be then stored in computer memory.
Should be appreciated that each several part of the present invention can realize with hardware, software, firmware or their combination.In the above-described embodiment, multiple step or method can with to store in memory and the software performed by suitable instruction execution system or firmware realize.Such as, if realized with hardware, the same in another embodiment, can realize by any one in following technology well known in the art or their combination: the discrete logic with the logic gates for realizing logic function to data-signal, there is the special IC of suitable combinational logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA) etc.
Those skilled in the art are appreciated that realizing all or part of step that above-described embodiment method carries is that the hardware that can carry out instruction relevant by program completes, described program can be stored in a kind of computer-readable recording medium, this program perform time, step comprising embodiment of the method one or a combination set of.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing module, also can be that the independent physics of unit exists, also can be integrated in a module by two or more unit.Above-mentioned integrated module both can adopt the form of hardware to realize, and the form of software function module also can be adopted to realize.If described integrated module using the form of software function module realize and as independently production marketing or use time, also can be stored in a computer read/write memory medium.
The above-mentioned storage medium mentioned can be ROM (read-only memory), disk or CD etc.Although illustrate and describe embodiments of the invention above, be understandable that, above-described embodiment is exemplary, can not be interpreted as limitation of the present invention, and those of ordinary skill in the art can change above-described embodiment within the scope of the invention, revises, replace and modification.
Claims (10)
1. a recommend method for information, is characterized in that, comprising:
Obtain the alternating content to be recommended of Query Information and correspondence thereof;
Obtain the parameter information of described Query Information and learning network corresponding to described alternating content to be recommended; And
Calculate described Query Information and the similarity of described alternating content to be recommended in pre-set space according to described parameter information, and from alternating content to be recommended, filter out content to be recommended according to described similarity, for representing described content to be recommended to user.
2. method according to claim 1, is characterized in that, describedly calculates described Query Information and the similarity of described alternating content to be recommended in pre-set space according to described parameter information, comprising:
Described Query Information primary vector corresponding in described pre-set space is calculated according to described parameter information;
Each alternating content to be recommended secondary vector corresponding in described pre-set space is calculated according to described parameter information; And
Calculate the similarity between described primary vector and each secondary vector.
3. method according to claim 2, is characterized in that, describedly from alternating content to be recommended, filters out content to be recommended according to described similarity, comprising:
Determine that described similarity is greater than the secondary vector of predetermined threshold value, and obtain described content to be recommended according to the secondary vector determined; Or
Determine the secondary vector of predetermined quantity according to described similarity order from high to low, and obtain described content to be recommended according to the secondary vector determined.
4. method according to claim 2, is characterized in that, the parameter information of the described Query Information of described acquisition and learning network corresponding to described alternating content to be recommended, comprising:
Corresponding learning network is built according to described Query Information and described alternating content to be recommended;
Obtain the training sample of described learning network, use described training sample to obtain described parameter information by preset algorithm.
5. method according to claim 4, is characterized in that, described preset algorithm comprises back-propagation algorithm and stochastic gradient descent method.
6. a recommendation apparatus for information, is characterized in that, comprising:
First obtains module, for obtaining the alternating content to be recommended of Query Information and correspondence thereof;
Second obtains module, for obtaining the parameter information of described Query Information and learning network corresponding to described alternating content to be recommended; And
Screening module, for calculating described Query Information and the similarity of described alternating content to be recommended in pre-set space according to described parameter information, and from alternating content to be recommended, filter out content to be recommended according to described similarity, for representing described content to be recommended to user.
7. device according to claim 6, is characterized in that, described screening module, specifically for:
Described Query Information primary vector corresponding in described pre-set space is calculated according to described parameter information;
Each alternating content to be recommended secondary vector corresponding in described pre-set space is calculated according to described parameter information; And
Calculate the similarity between described primary vector and each secondary vector.
8. device according to claim 7, is characterized in that, described screening module, specifically for:
Determine that described similarity is greater than the secondary vector of predetermined threshold value, and obtain described content to be recommended according to the secondary vector determined; Or
Determine the secondary vector of predetermined quantity according to described similarity order from high to low, and obtain described content to be recommended according to the secondary vector determined.
9. device according to claim 7, is characterized in that, described second obtains module, specifically for:
Corresponding learning network is built according to described Query Information and described alternating content to be recommended;
Obtain the training sample of described learning network, use described training sample to obtain described parameter information by preset algorithm.
10. device according to claim 9, is characterized in that, described preset algorithm comprises back-propagation algorithm and stochastic gradient descent method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410738395.5A CN104572825B (en) | 2014-12-04 | 2014-12-04 | The recommended method and device of information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410738395.5A CN104572825B (en) | 2014-12-04 | 2014-12-04 | The recommended method and device of information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104572825A true CN104572825A (en) | 2015-04-29 |
| CN104572825B CN104572825B (en) | 2019-03-12 |
Family
ID=53088887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410738395.5A Active CN104572825B (en) | 2014-12-04 | 2014-12-04 | The recommended method and device of information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104572825B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915860A (en) * | 2015-06-10 | 2015-09-16 | 无线生活（杭州）信息科技有限公司 | Commodity recommendation method and device |
| CN105045818A (en) * | 2015-06-26 | 2015-11-11 | 腾讯科技（深圳）有限公司 | Picture recommending method, apparatus and system |
| CN105809473A (en) * | 2016-02-29 | 2016-07-27 | 北京百度网讯科技有限公司 | Training method, service recommending method for coupling model parameters and associated apparatus |
| WO2016206188A1 (en) * | 2015-06-23 | 2016-12-29 | 百度在线网络技术（北京）有限公司 | Data processing method, device and apparatus for output object and non-volatile computer storage medium |
| CN106557554A (en) * | 2016-11-04 | 2017-04-05 | 北京百度网讯科技有限公司 | Display packing and device based on the Search Results of artificial intelligence |
| CN108388630A (en) * | 2018-02-13 | 2018-08-10 | 北京奇虎科技有限公司 | A kind of shopping information method for pushing, device and electronic equipment |
| CN108388629A (en) * | 2018-02-13 | 2018-08-10 | 北京奇虎科技有限公司 | A kind of multimedia information push method, device and electronic equipment |
| CN108399232A (en) * | 2018-02-13 | 2018-08-14 | 北京奇虎科技有限公司 | A kind of information-pushing method, device and electronic equipment |
| CN108415996A (en) * | 2018-02-13 | 2018-08-17 | 北京奇虎科技有限公司 | A kind of news information method for pushing, device and electronic equipment |
| CN109344299A (en) * | 2018-11-12 | 2019-02-15 | 考拉征信服务有限公司 | Object search method, apparatus, electronic equipment and computer readable storage medium |
| CN110162689A (en) * | 2018-05-10 | 2019-08-23 | 腾讯科技（北京）有限公司 | Information-pushing method, device, computer equipment and storage medium |
| CN110941766A (en) * | 2019-12-10 | 2020-03-31 | 北京字节跳动网络技术有限公司 | Information pushing method and device, computer equipment and storage medium |
| CN111625707A (en) * | 2020-05-29 | 2020-09-04 | 北京字节跳动网络技术有限公司 | Recommendation response method, device, medium and equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090259651A1 (en) * | 2008-04-11 | 2009-10-15 | Microsoft Corporation | Search results ranking using editing distance and document information |
| CN103353872A (en) * | 2013-06-03 | 2013-10-16 | 大连理工大学 | Teaching resource personalization recommendation method based on neural network |
-
2014
- 2014-12-04 CN CN201410738395.5A patent/CN104572825B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090259651A1 (en) * | 2008-04-11 | 2009-10-15 | Microsoft Corporation | Search results ranking using editing distance and document information |
| CN103353872A (en) * | 2013-06-03 | 2013-10-16 | 大连理工大学 | Teaching resource personalization recommendation method based on neural network |
Non-Patent Citations (3)
| Title |
|---|
| FANGXIANG FENG ET AL: "Cross-modal Retrieval with Correspondence Autoencoder", 《INTERNATIONAL CONFERENCE ON MULTIMEDIA》 * |
| 殷力昂: "一种在深度结构中学习原型的分类方法", 《中国优秀硕士学位论文全文数据库》 * |
| 王宁: "一种基于BP神经网络的即时在线推荐系统", 《计算机技术与发展》 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104915860A (en) * | 2015-06-10 | 2015-09-16 | 无线生活（杭州）信息科技有限公司 | Commodity recommendation method and device |
| WO2016206188A1 (en) * | 2015-06-23 | 2016-12-29 | 百度在线网络技术（北京）有限公司 | Data processing method, device and apparatus for output object and non-volatile computer storage medium |
| CN105045818A (en) * | 2015-06-26 | 2015-11-11 | 腾讯科技（深圳）有限公司 | Picture recommending method, apparatus and system |
| CN105809473A (en) * | 2016-02-29 | 2016-07-27 | 北京百度网讯科技有限公司 | Training method, service recommending method for coupling model parameters and associated apparatus |
| CN105809473B (en) * | 2016-02-29 | 2021-11-23 | 北京百度网讯科技有限公司 | Training method for matching model parameters, service recommendation method and corresponding device |
| CN106557554A (en) * | 2016-11-04 | 2017-04-05 | 北京百度网讯科技有限公司 | Display packing and device based on the Search Results of artificial intelligence |
| CN106557554B (en) * | 2016-11-04 | 2019-11-08 | 北京百度网讯科技有限公司 | The display methods and device of search result based on artificial intelligence |
| CN108415996A (en) * | 2018-02-13 | 2018-08-17 | 北京奇虎科技有限公司 | A kind of news information method for pushing, device and electronic equipment |
| CN108399232A (en) * | 2018-02-13 | 2018-08-14 | 北京奇虎科技有限公司 | A kind of information-pushing method, device and electronic equipment |
| CN108388629A (en) * | 2018-02-13 | 2018-08-10 | 北京奇虎科技有限公司 | A kind of multimedia information push method, device and electronic equipment |
| CN108388630A (en) * | 2018-02-13 | 2018-08-10 | 北京奇虎科技有限公司 | A kind of shopping information method for pushing, device and electronic equipment |
| CN110162689A (en) * | 2018-05-10 | 2019-08-23 | 腾讯科技（北京）有限公司 | Information-pushing method, device, computer equipment and storage medium |
| CN109344299A (en) * | 2018-11-12 | 2019-02-15 | 考拉征信服务有限公司 | Object search method, apparatus, electronic equipment and computer readable storage medium |
| CN110941766A (en) * | 2019-12-10 | 2020-03-31 | 北京字节跳动网络技术有限公司 | Information pushing method and device, computer equipment and storage medium |
| CN110941766B (en) * | 2019-12-10 | 2023-10-20 | 北京字节跳动网络技术有限公司 | Information pushing method, device, computer equipment and storage medium |
| CN111625707A (en) * | 2020-05-29 | 2020-09-04 | 北京字节跳动网络技术有限公司 | Recommendation response method, device, medium and equipment |
| CN111625707B (en) * | 2020-05-29 | 2023-04-14 | 北京字节跳动网络技术有限公司 | Recommendation response method, device, medium and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104572825B (en) | 2019-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104572825A (en) | Method and device for recommending information | |
| CN110457452B (en) | Recommendation reason generation method and device, electronic equipment and readable storage medium | |
| CN110737778B (en) | Knowledge graph and Transformer based patent recommendation method | |
| CN104239495A (en) | Search method and search device | |
| CN103353872B (en) | A kind of teaching resource personalized recommendation method based on neutral net | |
| CN108287864A (en) | A kind of interest group division methods, device, medium and computing device | |
| EP4181026A1 (en) | Recommendation model training method and apparatus, recommendation method and apparatus, and computer-readable medium | |
| CN106777229A (en) | Personalized recommendation real-time testing method and device and electronic equipment | |
| CN111652378B (en) | Learning to select vocabulary for category features | |
| CN104090963A (en) | Search information recommendation method and device | |
| CN104881484A (en) | Information recommendation method and device | |
| CN103646341B (en) | A kind of website provides the recommendation method and apparatus of object | |
| CN104809195A (en) | Recommending method and recommending device of search result | |
| CN110598095B (en) | Method, device and storage medium for identifying article containing specified information | |
| CN106445971A (en) | Application recommendation method and system | |
| CN110473073A (en) | The method and device that linear weighted function is recommended | |
| CN105701516B (en) | A kind of automatic image marking method differentiated based on attribute | |
| CN110781377B (en) | Article recommendation method and device | |
| Wu | A combined SR-CUSUM procedure for detecting common changes in panel data | |
| CN116401542A (en) | Multi-intention multi-behavior decoupling recommendation method and device | |
| CN115358809A (en) | Multi-intention recommendation method and device based on graph comparison learning | |
| CN103218460B (en) | Image tag complementing method based on the sparse reconstruct of optimum linearity | |
| CN114817552A (en) | Exercise association relation processing method, exercise association relation processing device, exercise association relation processing equipment and storage medium | |
| CN113129053A (en) | Information recommendation model training method, information recommendation method and storage medium | |
| CN103646053A (en) | Website providing object recommendation method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |