Invention content
This specification embodiment is intended to provide a kind of scheme of more effective gray scale publication new product, to solve the prior art
In deficiency.
To achieve the above object, this specification provides a kind of method of the publication of gray scale in systems new product on one side,
Including:Multiple keywords that multiple first products and the multiple first product include in acquisition system, wherein the multiple
First product includes new product to be released;Based on the multiple first product and the multiple first product include it is more
A keyword obtains the weight of each keyword of the new product;According to the weight, obtain the keyword of new product to
Amount;Calculate the similarity of the crucial term vector of the user obtained in advance and the crucial term vector of the new product;And described
In the case of similarity is more than predetermined threshold, determines that the user is the target user using the new product, new product is sent out
Cloth gives the target user.
In one embodiment, in the method that the gray scale in systems issues new product, the new product is old version
The more new version of this product, and determine that the user is to include using the target user of the new product, when the user into
When entering the early version product so that the user uses the new product.
In one embodiment, in the method that the gray scale in systems issues new product, the multiple first product
Including multiple keywords be multiple keywords that the multiple first product text that includes includes.
In one embodiment, in the method that the gray scale in systems issues new product, the new product is obtained
The weight of each keyword includes that the weight of each keyword of the new product is obtained by TF-IDF algorithms.
In one embodiment, in the method that the gray scale in systems issues new product, the use obtained in advance
The crucial term vector at family obtains as follows:Multiple second products and the multiple second product include in acquisition system
Multiple keywords, wherein second product relative to user open use;User is obtained about the multiple second product
Preference degrees of data;According to the preference degrees of data, the multiple second product is divided into positive sample set and negative sample collection
It closes;Based on multiple keywords that the multiple second product and the multiple second product include, each second product is obtained
Each keyword weight;According to the weight of each keyword of each second product, each second production is obtained
The crucial term vector of product;And the keyword according to the positive sample set, negative sample set and each second product
Vector is calculated by Rocchio algorithms, obtains the crucial term vector of user.
In one embodiment, in the method that the gray scale in systems issues new product, the multiple second product
It is multiple products that the multiple first product includes.
In one embodiment, in the method that the gray scale in systems issues new product, the preference data packet
Include at least one of following data:User is to the frequency of usage of product, user to the scoring of product and user to the recent of product
Access times.
Another aspect of the present invention provides a kind of device of the publication of gray scale in systems new product, including:First acquisition unit,
It is configured to, multiple keywords that multiple first products and the multiple first product include in acquisition system, wherein described more
A first product includes new product to be released;Second acquisition unit is configured to, and is based on the multiple first product and institute
Multiple keywords that multiple first products include are stated, the weight of each keyword of the new product is obtained;Third acquiring unit,
It is configured to, according to the weight, obtains the crucial term vector of new product;Computing unit is configured to, and calculates the user obtained in advance
Crucial term vector and the new product crucial term vector similarity;And determination unit, it is more than in the similarity pre-
In the case of determining threshold value, determines that the user is the target user using the new product, give the release of new products to the target
User.
In one embodiment, in the device that the gray scale in systems issues new product, the new product is old version
The more new version of this product, and, the determination unit is additionally configured to, when user enters the early version product so that institute
It states user and uses the new product.
In one embodiment, in the device that the gray scale in systems issues new product, the second acquisition unit
It is additionally configured to, the weight of each keyword of the new product is obtained by TF-IDF algorithms.
In one embodiment, the device of the publication of the gray scale in systems new product further includes the 4th acquiring unit, is matched
It is set to, obtains the crucial term vector of user in advance, the 4th acquiring unit specifically includes:First obtains subelement, is configured to,
Multiple keywords that multiple second products and the multiple second product include in acquisition system, wherein the second product phase
User is opened and is used;Second obtains subelement, is configured to, and obtains the preference number of degrees of the user about the multiple second product
According to;Divide subelement, be configured to, according to the preference degrees of data, by the multiple second product be divided into positive sample set and
Negative sample set;Third obtains subelement, is configured to, includes based on the multiple second product and the multiple second product
Multiple keywords, obtain the weight of each keyword of each second product;4th obtains subelement, is configured to, according to institute
The weight of each keyword of each second product is stated, the crucial term vector of each second product is obtained;And calculate son
Unit is configured to, and according to the crucial term vector of the positive sample set, negative sample set and each second product, is led to
It crosses Rocchio algorithms to be calculated, obtains the crucial term vector of user.
By issuing the scheme of new product according to the gray scale in systems of this specification embodiment, can be issued in gray scale new
Any active ues are effectively hit during product, so as to efficiently control gray scale rhythm, ensure gradation effect, and can be with
The problem of collecting new product in time simultaneously solves.
Specific implementation mode
This specification embodiment is described below in conjunction with attached drawing.
Fig. 1 shows the schematic diagram of the system 100 according to this specification embodiment.As shown in Figure 1, system 100 includes production
Product vector acquiring unit 11, user vector acquiring unit 12 and similarity calculated 13.In product vector acquiring unit 11
In, the keyword set that the first product set and multiple first product include is obtained, the multiple first product includes general
The new product for wanting gray scale to issue.Based on the multiple first product and the keyword set, each key of new product is obtained
The weight of word, and according to the weight, the crucial term vector of new product is obtained, to obtain the crucial term vector of new product, and
The crucial term vector of the new product is sent to similarity calculated 13.Meanwhile in product vector acquiring unit 11, obtain
Relative to user open the second product set for using and its including keyword set, similarly, relative to the second product collection
Conjunction and its keyword set, can obtain the weight of each keyword of each second product, to obtain each second product
Crucial term vector, and send it to user vector acquiring unit 12.In user vector acquiring unit 12, user is obtained to each
The preference degrees of data of a second product, so as to which the second product set is divided into positive sample set according to preference degrees of data and is born
Sample set.By Rocchio algorithms, it is based on the positive sample vector set, negative sample vector set and each second product
Crucial term vector, the crucial term vector of user can be calculated, and send it to similarity calculated 13.Similarity calculation
Unit 13 calculates the similarity between the crucial term vector and user's key term vector of the new product that it is received, when the similarity is big
In the case of predetermined threshold, it may be determined that the user is the gray scale target user using new product.
Fig. 2 shows the flow charts for the method that new product is issued according to the gray scale in systems of this specification embodiment.Institute
The method of stating includes:Multiple keys that multiple first products and the multiple first product include in step S21, acquisition system
Word, wherein the multiple first product includes new product to be released;In step S22, based on the multiple first product with
And multiple keywords that the multiple first product includes, obtain the weight of each keyword of the new product;In step
S23 obtains the crucial term vector of new product according to the weight;In step S24, the user in the system obtained in advance is calculated
Crucial term vector and the new product crucial term vector similarity;And in step S25, it is more than in the similarity pre-
In the case of determining threshold value, determines that the user is the target user using the new product, give the release of new products to the target
User.
First, multiple passes that multiple first products and the multiple first product include in step S21, acquisition system
Keyword, wherein the multiple first product includes new product to be released.The system for example can be the net in internet
It stands, the APP in terminal device, such as Alipay APP etc..The multiple first product for example can be the whole that APP includes
Existing product and will gray scale publication new product.For example, multiple first product includes N number of product, the set of N number of product
For:
D={ d1, d2..., dN}。
Its respective multiple keyword can be obtained from each product.Keyword is obtained in the text that can include from product.?
In one embodiment, the participle that can be used as keyword in product text is determined by product development personnel.In one embodiment,
Choose keyword of the larger participle of information content in product text as product.By obtaining multiple keywords of each product,
And brought together, the keyword set T of the multiple first product can be obtained.Such as it is wrapped in the keyword set T
Include n keyword:
T={ t1, t2..., tn}。
It is obtained based on multiple keywords that the multiple first product and the multiple first product include in step S22
Take the weight of each keyword of the new product.For example, power of i-th of keyword in new product j in keyword set T
Weight is wij.The weight wijAssignment can be carried out by a variety of methods.
In one embodiment, it is assumed that each keyword in new product j is of equal importance, then when new product j includes
When i-th of keyword, then wij=1, when not including i-th of keyword in new product j, then wij=0.For example, new product is foot
Mark 2.0 comprising keyword " credit, bill, behavior ", such as its be respectively in keyword set T first, second, and third
A keyword, then for new product j, w1j=1, w2j=1, w3j=1, and wij=0, wherein i=4 to n.
In one embodiment, each of the new product is obtained by TF-IDF algorithms shown in following formula (1) and (2)
The weight of a keyword.Wherein:
TF-IDF (i, j)=TF (i, j) * IDF (i) (1),
TF (i, j) wherein in formula (1) is normalized word frequency of the keyword i in product j, is passed through formula (3)
It calculates and obtains,
Wherein ni,jThe number occurred in product j for keyword i.
N (i) in formula (2) is the product number of keyword i occur in above-mentioned N number of product.By returning to TF-IDF (i, j)
One changes, and obtains wij, as shown in formula (4),
The weight that each keyword of the new product is obtained by above-mentioned TF-IDF algorithms, by keyword in product
The number (other product numbers for including the keyword) that number, the keyword of portion's appearance occur in other products determines keyword
Weight, that is, keyword is higher in the frequency of interiors of products, and weight is bigger, and keyword is bigger in the number that other products occur, then
Weight is smaller, to more accurately define the weight of each keyword of product.
The crucial term vector of new product is obtained according to the weight in step S23.For example, for new product j, obtaining
In above-mentioned keyword set after respective weight of the n keyword in product j, the crucial term vector of new product can be obtainedWhereinFor example, for the new product of the footprint 2.0 in step S22, its pass can be obtained
Keyword vector is
In step S24, the phase of the crucial term vector and the crucial term vector of the new product of the user obtained in advance is calculated
Like degree.
First, the acquisition to user's key term vector is illustrated with reference to figure 3.Fig. 3 is shown according to this specification embodiment
Obtain the flow chart of the method for user's key term vector.
As shown in figure 3, in step S31, multiple second products and the multiple second product include in acquisition system
Multiple keywords use wherein second product is opened relative to user.The multiple second product for example can be in APP
All over products in addition to new product.It is appreciated that the multiple second product is not limited to belong to the multiple first product
Product, as long as it is that opening uses, and can get user to its preference degrees of data in face of user.
In step S32, preference degrees of data of the user about the multiple second product is obtained.Wherein, the preference number of degrees
According to including at least one of following data:User is to the frequency of usage of product, user to the scoring of product and user to product
Recent access times.
In step S33, according to the preference degrees of data, the multiple second product is divided into positive sample set and negative sample
This set.The product set that the positive sample set, that is, user likes, the product collection that the negative sample set, that is, user does not like
It closes.In one embodiment, when user is to the frequency of usage (such as the number used daily or the number used weekly) of product
When more than predetermined threshold, determine that user likes the product.In one embodiment, when user is more than predetermined point to the scoring of product
When value, determine that user likes the product.In one embodiment, when user to product it is recent (in such as 2 days, in 3 days, one week
It is interior etc.) access times when being more than or equal to pre-determined number, determine that user likes the product, for example, making to product in 3 days as user
When being more than or equal to 1 time with number, determine that user likes the product.In one embodiment, consider every preference degrees of data,
To determine preference of the user to product.
It is obtained based on multiple keywords that the multiple second product and the multiple second product include in step S34
Take the weight of each keyword of each second product;In step S35, according to each keyword of each second product
Weight obtains the crucial term vector of each second product.Here, step S34 and step S35 and step described in reference diagram 2
Rapid S22 and S23 is essentially identical, and details are not described herein.
In step S36, according to the keyword of the positive sample set, negative sample set and each second product to
Amount obtains the crucial term vector of user by Rocchio algorithms.Wherein, according to Rocchio algorithms, pass through following formula (5)
Obtain the crucial term vector of user
Wherein IrFor positive sample set, InrFor negative sample set,To belong to the key of the product in positive sample set
Term vector,To belong to the crucial term vector of the product in negative sample set, and, β and the power that γ is positive and negative sample set
Weight, size are determined by system according to the distribution situation of positive negative sample.
The method of above-mentioned acquisition user's key term vector described in reference diagram 3 can be periodically executed, such as once per week,
Or execute daily once, the crucial term vector of user is constantly updated to the service condition of product in APP according to user.This is obtained
The method for taking family key term vector can also in real time be carried out when needing using user's key term vector, for example, implementing root
When issuing the method for new product according to the gray scale of this specification embodiment, the crucial term vector of user is calculated in real time, so as to carry
For more accurate user vector.
Fig. 2 is returned, in the crucial term vector for obtaining user as described aboveLater, can calculate the keyword of user to
AmountWith the crucial term vector of above-mentioned new productBetween similarity.
Can calculate the similarity between two vectors in several ways, for example, Euclidean distance, manhatton distance, Ming Shi away from
From, cosine similarity etc..It is preferred that calculating the crucial term vector of user by following cosine similarity formula (6)With it is upper
State the crucial term vector of new productBetween similarity.
The value range of the similarity is [- 1,1], and when its value is closer to 1, two vectors of expression are closer.Described
In the case of multiple second product set are not belonging to the multiple first product set, i.e. the crucial term vector of userWith it is upper
State the crucial term vector of new productBetween characteristic dimension may be different, can by the characteristic dimension lacked to vector into
Row mends 0 to calculate the similarity between two vectors.
In step S25, in the case of the similarity is more than predetermined threshold, determine that the user is to use the new production
The target user of product gives the release of new products to the target user.For example, in the case of by above-mentioned cosine similarity, it can
To set predetermined threshold as 0.9.When similarity is more than predetermined threshold 0.9, the crucial term vector and new product of user are indicated
Crucial term vector coincidence factor is higher, and the probability that user likes new product is larger, that is, user is larger using the probability of new product.Cause
This, determines that the user is the target user using new product, can show new product to the user in APP, can be by logical
The mode known invites the user to use the new product, etc..It thereby may be ensured that and gone all out to do one's duty regardless of personal danger during gray scale issues new product
Middle any active ues, to collect problem rapidly and to solve.
In one embodiment, the new product is the more new version of early version product, and, it is more than in the similarity
In the case of predetermined threshold, when user enters the early version product so that the user uses the new product.To,
So that the height of new product is using probability user, only when entering early version product, ability is more promoted by automatic shunt to new product
User hit rate.In an example, it is the update of " footprint 1.0 " by gray scale publication new product " footprint 2.0 " in APP
Version.It is " credit, bill, behavior " that the keyword that it includes is obtained from " footprint 2.0 " text.According to APP include it is more
A product and its including keyword, can get new product crucial term vector (1,1,1,0 ... 0)T.It is recently entered in user
In the case of footprint 1.0, according to the preference degrees of data, it may be determined that footprint 1.0 is the product of user liked, and footprint 1.0 includes
Keyword:Credit, bill, behavior.Therefore, according to above-mentioned formula (4), the crucial term vector of user includes being multiplied by weight
(1,1,1,0,…0)T.It can determine by formula (5), the crucial term vector of user and the keyword vector similarity of the new product
More than predetermined threshold, determine that user is the target user of new product.Therefore, when user is again introduced into footprint so that user makes
With footprint 2.0.
In one embodiment, the method for new product being issued according to the gray scale in systems of this specification embodiment can be regular
It executes, such as is executed daily once, according to newer user's key term vector, to judge whether user is changed into the mesh of new product
Mark user.For example, as described in example above, in the case of new product is the more new version of early version product, closed in user
In the case of keyword vector is less than predetermined threshold with new product keyword vector similarity, the target that user does not become new product is used
Family.Therefore, when user enters early version product, early version product is still used.But enter old version in the recent period in user
After this product, early version product is determined as the product that user likes by the preference degrees of data that system uses in the recent period according to this, and
Update the crucial term vector of user.To be issued according to the gray scale in systems of this specification embodiment when system executes again
The method of new product, and wherein use newer user's key term vector when, it may be determined that user has turned to new product
Target user.Therefore, when user is again introduced into early version product so that the user uses new product.
Fig. 4 shows a kind of device 400 of the publication of gray scale in systems new product.Device 400 includes:First acquisition unit
41, it is configured to, multiple keywords that multiple first products and the multiple first product include in acquisition system, wherein described
Multiple first products include new product to be released;Second acquisition unit 42, is configured to, based on the multiple first product with
And multiple keywords that the multiple first product includes, obtain the weight of each keyword of the new product;Third obtains
Unit 43, is configured to, and according to the weight, obtains the crucial term vector of new product;Computing unit 44, is configured to, and calculates advance
The similarity of the crucial term vector of user in the system of acquisition and the crucial term vector of the new product;And determination unit
45, in the case of the similarity is more than predetermined threshold, determine that the user is the target user using the new product, it will
The target user is given in the release of new products.
In one embodiment, the new product is the more new version of early version product, and, the determination unit is also matched
It is set to, when user enters the early version product so that the user uses the new product.
In one embodiment, the second acquisition unit is additionally configured to, and the new product is obtained by TF-IDF algorithms
Each keyword weight.
In one embodiment, described device 400 further includes the 4th acquiring unit 46, is configured to, and obtains user's in advance
Crucial term vector.4th acquiring unit 46 specifically includes:First obtains subelement 461, is configured to, multiple in acquisition system
Multiple keywords that second product and the multiple second product include make wherein second product is opened relative to user
With;Second obtains subelement 462, is configured to, and obtains preference degrees of data of the user about the multiple second product;It is single to divide son
Member 463, is configured to, according to the preference degrees of data, the multiple second product is divided into positive sample set and negative sample collection
It closes;Third obtain subelement 464, be configured to, based on the multiple second product and the multiple second product include it is more
A keyword obtains the weight of each keyword of each second product;4th obtains subelement 465, is configured to, according to described
The weight of each keyword of each second product obtains the crucial term vector of each second product;And calculate son list
Member 466, is configured to, and according to the crucial term vector of the positive sample set, negative sample set and each second product, leads to
It crosses Rocchio algorithms to be calculated, obtains the crucial term vector of user.
By issuing the scheme of new product according to the gray scale in systems of this specification embodiment, can be issued in gray scale new
Any active ues are effectively hit during product, so as to efficiently control gray scale rhythm, ensure gradation effect, and can be with
The problem of collecting new product in time simultaneously solves.
Those of ordinary skill in the art should further appreciate that, be described in conjunction with the embodiments described herein
Each exemplary unit and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clear
Illustrate to Chu the interchangeability of hardware and software, generally describes each exemplary group according to function in the above description
At and step.These functions hold track with hardware or software mode actually, depending on technical solution specific application and set
Count constraints.Those of ordinary skill in the art can be described to be realized using distinct methods to each specific application
Function, but this realization is it is not considered that exceed scope of the present application.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can hold track with hardware, processor
Software module or the combination of the two implement.Software module can be placed in random access memory (RAM), memory, read-only storage
Device (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology neck
In any other form of storage medium well known in domain.
Above-described specific implementation mode has carried out further the purpose of the present invention, technical solution and advantageous effect
It is described in detail, it should be understood that the foregoing is merely the specific implementation mode of the present invention, is not intended to limit the present invention
Protection domain, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.