CN103577473A

CN103577473A - Classification and disambiguation method, classification and disambiguation device and system thereof

Info

Publication number: CN103577473A
Application number: CN201210276178.XA
Authority: CN
Inventors: 黄申; 韩军
Original assignee: Niuhai Information Technology (Shanghai) Co Ltd
Current assignee: Beijing Jingdong Shangke Information Technology Co Ltd
Priority date: 2012-08-03
Filing date: 2012-08-03
Publication date: 2014-02-12
Anticipated expiration: 2032-08-03
Also published as: CN103577473B

Abstract

The invention discloses a classification and disambiguation method, a classification and disambiguation device and a system thereof. The classification and disambiguation method comprises the following steps: S1, reading in a keyword; S2, searching for product names obtained by retrieval of the keyword by a user, including the clicking rate and purchasing rate of each product type of the keyword; S3, calculating the relevance coefficient of each product type via the clicking probability and purchasing probability of the product type in sequence; S4, extracting a product type of which the number is equal to a displayed product number from the product type for serving as a displayed product type; S5, extracting the product picture of any of the products included in each displayed product type in sequence from a product database for serving as the product picture of the displayed product type; S6, displaying the product name and product picture of the displayed product type. The potential needs of all users are dug and matched via the clicking and purchasing behaviors of the users on a keyword search result, so that the accuracy and reliability of the search result are increased.

Description

Classification disappear qi method, classification disappear qi device and system thereof

Technical field

The present invention relates to a kind of classification disappear qi method, classification disappear qi device and system thereof, particularly relate to a kind of by the product classification obtaining by key search disappear qi method, classification disappear qi device and system thereof.

Background technology

Along with Internet era arrival, the ecommerce of China has obtained vigorous growth in recent years, main merchandising also, from books class commodity and electrical type commodity the earliest, develops into present household consumption product and popular general merchandise etc.And the type of merchandize of rapid expanding also makes product search play the part of more and more important role in e-commerce field.

Meanwhile, product search is also faced with a lot of new challenges.There is ambiguity in the key word of the inquiry of many times user's input, for example, while inputting key word " apple ", user's expectation may be both electronic product, also may be fruit, also for example, during input key word " Philip ", both may need shaver, also may need hair dryer etc., so generation of ambiguity, making to have in the result of user search is not the result that user needs in a large number, thereby makes whole Search Results inaccurate, thereby has suppressed to a great extent the action of user search product.

Summary of the invention

The technical problem to be solved in the present invention is in order to overcome the inaccurate and insecure defect of Search Results that in prior art, in product search, ambiguity key word causes, a kind of classification disappear qi method, classification disappear qi device and system thereof are provided, by all users the click of keyword search result and buying behavior are excavated to the Search Results of match user potential demand, thereby improve accuracy and the reliability of Search Results.

The present invention solves above-mentioned technical matters by following technical proposals:

The invention provides a kind of classification qi method that disappears, the wherein said classification qi method that disappears comprises the following steps:

S ₁, read in a key word;

S ₂, from user's click logs, search the click ratio of each product category that name of product that user obtains described key search comprises described key word, and buy from user the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record;

S ₃, click probability and the purchase probability by product category described in each calculates described product category successively relative coefficient;

S ₄, from described product category, extract product category that quantity equals a demonstration product number as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category;

S ₅, in a product database, successively each is shown to the product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category;

S ₆, show name of product and the product picture of described demonstration product category.

Wherein said product database is the product related informations such as name of product, product category and product picture of having stored specific product in prior art.And described product database can be retrieved based on product related informations such as described name of product, product category and product pictures respectively.

In the click logs of user described in the present invention, stored when a key word is retrieved the probability of the click of the different product category that the name of product that all users that retrieved described key word obtain retrieval comprises described key word.Wherein described in the present invention, clicking probability calculates by following formula:

CP (i | q) = \frac{click (i, q)}{freq (q)}

Wherein q is the key word of inquiry, and described i is product category, and described click (i, q) is the number of times that key word of the inquiry q clicks product category i later, and freq (q) is the number of times of key word of the inquiry q.

Same described user buys in daily record and has stored when a key word is retrieved, the probability of the purchase of the different product category that the name of product that all users that retrieved described key word obtain retrieval comprises described key word.

Purchase probability calculates by following formula:

PP (i | q) = \frac{purchase (i, q)}{freq (q)}

Wherein q is the key word of inquiry, and described i is product category, and described purchase (i, q) is the number of times that key word of the inquiry q buys product category i later, and freq (q) is the number of times of key word of the inquiry q.

In the present invention, utilize having retrieved the probability of click of the different product category that the user of described key word comprises described key word to name of product and the product that the probability of purchase is confirmed the needs that user is potential, so the probability of click of the different product category that user comprises described key word to name of product and the Data Collection of the probability of purchase are not problem to be solved by this invention, and the probability of click of different product categories and the collection of the probability of purchase that described user comprises described key word to name of product can obtain by the combination of the hardware and softwares such as data acquisition and processing (DAP) method of the prior art, so be no longer described in detail in the present invention.

Probability calculation in the probability by described click and purchase obtains after relative coefficient, in the present invention, according to the size order of described relative coefficient, from the corresponding product category of maximum relative coefficient, start to extract, until the quantity of the product category extracting and described demonstration product are counted location of equal.

Preferably, described step S ₃for:

Described product category successively through type 1 calculates the relative coefficient of described product category:

Relevance=w ₁* CP (i|q)+w ₂* PP (i|q) formula 1

Wherein said relevance is relative coefficient, described CP(i|q) for clicking probability, described PP(i|q) be purchase probability, described w ₁for clicking probability right coefficient, described w ₂for purchase probability weight coefficient; Described purchase probability weight coefficient is greater than described click probability right coefficient.

The degree of the potential purchasing demand of user that can represent due to buying behavior and click behavior is different, so the present invention regulates described buying behavior and the influence degree of click behavior to user's potential demand by purchase probability weight coefficient and click probability right coefficient, thereby potential demand of better digging user.

And concrete buying behavior more can the potential purchasing demand of representative of consumer with respect to click behavior, so think in the present invention that described purchase probability weight coefficient is greater than described click probability right coefficient.

Preferably, described purchase probability weight coefficient w ₁be 0.7, described click probability right coefficient w ₂be 0.3.

Preferably, described step S ₄for:

Described product category, according to the size of relative coefficient, is arranged as to a product category sequence according to order from big to small;

In described product category sequence, from thering is the product category of maximal correlation property coefficient, start to extract product category that quantity equals described demonstration product number as described demonstration product category.

In the present invention, preferably described product category can be sorted according to relative coefficient, thereby obtain several product categories of relative coefficient maximum, thereby obtain the product category of match user potential demand.

Preferably, described step S ₄for:

S ₄₁, from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category;

S ₄₂, the numerical value of described demonstration product number is reduced to 1, whether and to detect described demonstration product number be zero, if enter step S ₅, otherwise return to step S ₄₁.

Equally, the present invention can also preferably adopt the mode of traversal circulation to determine the product category of match user potential demand.

Preferably, described demonstration product number is 4,5 or 6.

In the present invention, consider the constraints such as final display effect, so showing that product number is 4,5 or 6 can realize best prompting effect.

Preferably, described step S ₅for:

In described product database, retrieve successively all products that comprise described demonstration product category, and between described product and described demonstration product category, establish the link the product picture using the product picture of any one product in the described product that comprises described demonstration product category as described demonstration product category.

The product picture of the present invention using the product picture in corresponding any one specific product of described demonstration product category as described demonstration product category, thus can be automatically that described demonstration product category is set product picture.

Preferably, described step S ₅for:

In described product database, retrieve successively all products that comprise described demonstration product category, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

Further, the product picture of the present invention using the product picture of first product after the corresponding all product sequences of described demonstration product category as described demonstration product category, automatically sets the ability of product picture thereby improved further for described demonstration product category.

Preferably, described step S ₆for:

In a viewing area, build a recommendation viewing area, and successively name of product and the product picture of each demonstration product category are shown in described recommendation viewing area according to order from left to right.

Preferably, described step S ₆in further comprising the steps of:

The name of product of described demonstration product category is shown in the below of the product picture of described demonstration product category.

The mode of the present invention by product picture and name of product combination pointed out recommended products classification for user, thereby is convenient to searching and selecting of user.

The present invention also provides a kind of classification qi method that disappears, and the wherein said classification qi method that disappears comprises the following steps:

S ₁₀₁, read in a key word;

S ₁₀₂, from user's click logs, search first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user;

S ₁₀₃, by first of product category described in each, click the first relevance parameter that probability and the first purchase probability calculate described product category successively;

S ₁₀₄, from user's click logs, search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user;

S ₁₀₅, by second of product category described in each, click the second relevance parameter that probability and the second purchase probability calculate described product category successively;

S ₁₀₆, the relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter successively;

S ₁₀₇, from described product category, extract product category that quantity equals a demonstration product number as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category;

S ₁₀₈, in a product database, successively each is shown to the product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category;

S ₁₀₉, show name of product and the product picture of described demonstration product category.

Because the seasonality of some products is more intense.For example the dress ornament in summer and winter just completely different, user's demand also can great changes have taken place.If just only adopt click volume and the sales volume in whole selling time section, have very large deviation when transform season.The present invention is by reducing further the deviation of seasonal product to the click behavior of the product category that comprises described key word and buying behavior to user in certain hour.

Preferably, described the first Preset Time section is less than the second Preset Time section.

In the present invention, by the neutralization of a short time period and a long-time section, consider the deviation of calming down seasonal product click and purchase probability.

Preferably, described the first Preset Time is 1 week, and described the second Preset Time is 1 month.

Preferably, described step S ₁₀₃for:

Described product category successively through type 2 calculates the first relevance parameter of described product category:

Relevance ₁=w ₁₁* CP ₁(i|q)+w ₁₂* PP ₁(i|q) formula 2

Wherein said relevance ₁be the first relevance parameter, described CP ₁(i|q) be the first click probability, described PP ₁(i|q) be the first purchase probability, described w ₁₁be the first click probability right coefficient, described w ₁₂it is the first purchase probability weight coefficient; Described the first purchase probability weight coefficient is greater than described first and clicks probability right coefficient.

Preferably, described the first purchase probability weight coefficient w ₁₁be 0.7, described first clicks probability right coefficient w ₁₂be 0.3.

Preferably, described step S ₁₀₅for:

Described product category successively through type 3 calculates the second relevance parameter of described product category:

Relevance ₂=w ₂₁* CP ₂(i|q)+w ₂₂* PP ₂(i|q) formula 3

Wherein said relevance ₂be the second relevance parameter, described CP ₂(i|q) be the second click probability, described PP ₂(i|q) be the second purchase probability, described w ₂₁be the second click probability right coefficient, described w ₂₂it is the second purchase probability weight coefficient; Described the second purchase probability weight coefficient is greater than described second and clicks probability right coefficient.

Preferably, described the second purchase probability weight coefficient w ₂₁be 0.7, described second clicks probability right coefficient w ₂₂be 0.3.

Preferably, described step S ₁₀₆for:

Described product category successively through type 4 calculates the relative coefficient of described product category:

Relevance_optimized=w ₃* relevance ₁+ w ₄* relevance ₂formula 4

Wherein said relevance ₁be the first relevance parameter, described relevance ₂be the second relevance parameter, described w ₃be the first relevance parameter weight coefficient, described w ₄be the second relevance parameter weight coefficient, wherein said the first relevance parameter weight coefficient is greater than described the second relevance parameter weight coefficient.

Because the data of nearest 1 week more can reflect calendar variation than 1 month, so the present invention is set as the first relevance parameter weight coefficient that has characterized short time period to be greater than the second relevance parameter weight coefficient that has characterized long-time section, thus the click and the excavation of buying behavior on the impact of user's potential demand that improve product in short time period.

Preferably, described step S ₁₀₇for:

S ₁₀₇₁, from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category;

S ₁₀₇₂, the numerical value of described demonstration product number is reduced to 1, whether and to detect described demonstration product number be zero, if enter step S ₁₀₈, otherwise return to step S ₁₀₇₁.

Preferably, described demonstration product number is 4,5 or 6.

Preferably, described step S ₁₀₈for:

Preferably, described step S ₁₀₉for:

Preferably, described step S ₁₀₉in further comprising the steps of:

The present invention provides again a kind of classification qi device that disappears, and is characterized in that the described classification qi device that disappears comprises:

One reads in module, for reading in a key word;

One retrieval module, for search the click ratio of each product category that name of product that user obtains described key search comprises described key word from user's click logs, and buy from user the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record;

One processing module, for the relative coefficient that click probability and the purchase probability by product category described in each calculates described product category successively;

One screening module, for extracting product category that quantity equals a demonstration product number from described product category as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category;

One picture link module, the product picture for the product picture of arbitrary product in a product database shows each included product of product category successively as described demonstration product category;

One display module, for showing name of product and the product picture of described demonstration product category.

Preferably, in product category described in described processing module successively through type 5, calculate the relative coefficient of described product category:

Relevance=w ₁* CP (i|q)+w ₂* PP (i|q) formula 5

Preferably, described screening module comprises:

One order module, for by described product category according to the size of relative coefficient, according to order from big to small, be arranged as a product category sequence;

One extraction module, in described product category sequence, starts to extract product category that quantity equals described demonstration product number as described demonstration product category from having the product category of maximal correlation property coefficient.

Preferably, described screening module comprises:

One searches module, for from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category;

One counting module, whether for the numerical value of described demonstration product number is reduced to 1, and to detect described demonstration product number be zero, if described picture link module executable operations, otherwise search module described in again calling.

Preferably, described demonstration product number is 4,5 or 6.

Preferably, described picture link module comprises that one for retrieving all products that comprise described demonstration product category at described product database successively, and between described product and described demonstration product category, establish the link the module using the product picture of any one product in the described product that comprises described demonstration product category as the product picture of described demonstration product category.

Preferably, described picture link module is for retrieving all products that comprise described demonstration product category at described product database successively, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

Preferably, described display module is used for building a recommendation viewing area in a viewing area, and successively name of product and the product picture of each demonstration product category is shown in described recommendation viewing area according to order from left to right.

Preferably, in described display module, also comprise that one for being shown in the name of product of described demonstration product category the module of below of the product picture of described demonstration product category.

The present invention separately provides a kind of classification qi device that disappears, and is characterized in that the described classification qi device that disappears comprises:

One reads in module, for reading in a key word;

One first retrieval module, for search first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word from user's click logs, click ratio, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user;

One first processing module, for clicking by first of product category described in each the first relevance parameter that probability and the first purchase probability calculate described product category successively;

One second retrieval module, for search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word from user's click logs, click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user;

One second processing module, for clicking by second of product category described in each the second relevance parameter that probability and the second purchase probability calculate described product category successively;

One the 3rd processing module, for the relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter successively;

Preferably, product category described in described the first processing module successively through type 6 calculate the first relevance parameter of described product category:

Relevance ₁=w ₁₁* CP ₁(i|q)+w ₁₂* PP ₁(i|q) formula 6

Preferably, product category described in described the second processing module successively through type 7 calculate the second relevance parameter of described product category:

Relevance ₂=w ₂₁* CP ₂(i|q)+w ₂₂* PP ₂(i|q) formula 7

Preferably, product category described in described the 3rd processing module successively through type 8 calculate the relative coefficient of described product category:

Relevance_optimized=w ₃* relevance ₁+ w ₄* relevance ₂formula 8

Preferably, described screening module comprises:

Preferably, described demonstration product number is 4,5 or 6.

The present invention separately provides a kind of classification qi system that disappears, and is characterized in that the described classification qi system that disappears comprises:

One storer, wherein stores a product database;

One input media, for reading in a key word;

One server, described server comprises that user's click logs and a user buy daily record, wherein said server is searched the click ratio of each product category that name of product that user obtains described key search comprises described key word from user's click logs, and buy the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record from user, and the relative coefficient that click probability and the purchase probability by product category described in each calculates described product category successively;

One processor, for extracting product category that quantity equals a demonstration product number from described product category as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category, and in the product database of described storer, successively each is shown to the product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category;

One display, for showing name of product and the product picture of described demonstration product category.

The present invention also provides a kind of classification qi system that disappears, and is characterized in that the described classification qi system that disappears comprises:

One storer, wherein stores a product database;

One input media, for reading in a key word;

One server, described server comprises that user's click logs and a user buy daily record, wherein said server is searched first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word and is clicked ratio from user's click logs, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user, and from user's click logs, search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user, and described server also clicks by first of product category described in each the first relevance parameter that probability and the first purchase probability calculate described product category successively, and by second of product category described in each, click the second relevance parameter that probability and the second purchase probability calculate described product category successively, then the relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter successively,

Positive progressive effect of the present invention is:

Classification of the present invention disappear qi method, classification disappear qi device and system thereof, by all users the click of keyword search result and buying behavior are excavated to the Search Results of match user potential demand, thereby improve accuracy and the reliability of Search Results.

In addition the present invention also excavates coupling to all users in different time intervals to the click of keyword search result and buying behavior by limiting, thereby lower the error to the search of seasonal product, improve further accuracy and the reliability of Search Results.

Accompanying drawing explanation

Fig. 1 is the disappear process flow diagram of qi method of the classification of the first embodiment of the present invention.

Fig. 2 is the disappear process flow diagram of qi method of the classification of the second embodiment of the present invention.

Fig. 3 is the classification of the third embodiment of the present invention qi device that disappears.

Fig. 4 is the classification of the fourth embodiment of the present invention qi device that disappears.

Fig. 5 is the classification of the fifth embodiment of the present invention qi device that disappears.

Fig. 6 is the classification of the sixth embodiment of the present invention qi device that disappears.

Fig. 7 is the classification of the seventh embodiment of the present invention qi system that disappears.

Embodiment

Below in conjunction with accompanying drawing, provide preferred embodiment of the present invention, to describe technical scheme of the present invention in detail.

The first embodiment:

The classification of the present embodiment disappear qi method by all users the click of keyword search result and buying behavior are excavated to the Search Results of match user potential demand, thereby improve accuracy and the reliability of Search Results.According to the statistics of experimental data, the classification by the present embodiment disappears the demonstration product category that qi method calculates can more than 85% screening requirements of encompasses users.

Wherein as shown in Figure 1, the classification of the present embodiment qi method that disappears comprises the following steps:

Step 11, reads in a key word.Wherein key word described in the present embodiment is the name of product of user's input etc., such as " apple " or " Philip " etc.

Step 12, from user's click logs, search the click ratio of each product category that name of product that user obtains described key search comprises described key word, and buy from user the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record.

In the click logs of user described in the present embodiment, stored when a key word is retrieved the probability of the click of the different product category that the name of product that all users that retrieved described key word obtain retrieval comprises described key word.Wherein described in the present embodiment, clicking probability calculates by following formula:

CP (i | q) = \frac{click (i, q)}{freq (q)}

Purchase probability calculates by following formula:

PP (i | q) = \frac{purchase (i, q)}{freq (q)}

Step 13, described product category is through type 1 and the click probability of product category described in each and the relative coefficient that purchase probability calculates described product category successively.

Relevance=w ₁* CP (i|q)+w ₂* PP (i|q) formula 1

Concrete buying behavior more can the potential purchasing demand of representative of consumer with respect to click behavior, so think in the present invention that described purchase probability weight coefficient is greater than described click probability right coefficient.Wherein said purchase probability weight coefficient and the concrete data of described click probability right coefficient can regulate according to specific needs, preferably set w in the present embodiment ₁=0.3, w ₂=0.7, now there is best excavation matching effect.

For example, user entered keyword is " A ", and described key word A relates to product category A1, A2 and A3, and sets w ₁=0.3, w ₂=0.7 o'clock, after input " A ", obtaining user, to click A1 probability be 30%, and the probability that user clicks A2 is 30%, and the probability that user clicks A3 is 40%.The probability that user buys A1 is 20%, and the probability that user buys A2 is 50%, and the probability that user buys A2 is 30%.Being calculated as follows of relative coefficient now:

relevance(A1)=0.3×0.3+0.7×0.2=0.23

relevance(A2）=0.3×0.3+0.7×0.5=0.44

relevance(A3）=0.3×0.4+0.7×0.3=0.33

Step 14, according to the size of relative coefficient, is arranged as a product category sequence according to order from big to small by described product category.For example the relative coefficient of the said goods classification A1, A2 and A3 is sorted, now the order of described product category sequence is relevance(A2), relevance(A3) and relevance(A1).

Step 15, in described product category sequence, starts to extract product category that quantity equals described demonstration product number as described demonstration product category from having the product category of maximal correlation property coefficient.

For example, if described demonstration product number is, from described product category sequence, take out relevance(A2 at 2 o'clock), relevance(A3).

In addition in above-mentioned steps 14 and step 15, can also adopt the mode of searching loop to take out correspondingly relative coefficient.

For example, first from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category, then the numerical value of described demonstration product number is reduced to 1, and whether detect described demonstration product number be zero, if enter step 16, otherwise repeat this step.

Because actual product kind is thousands of, so can comprise a lot of product categories in described product category sequence, but consider the eye-catching object of restriction and the demonstration of actual displayed, described in the present embodiment, show that product number can choose any in 4,5 or 6, wherein preferably choose 5, many demonstration product categories of now not only can trying one's best, can also take into account the highlighting of demonstration.

Step 16, shows each that in a product database product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category successively.

And in the present embodiment, in described product database, retrieve successively all products that comprise described demonstration product category, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

Be the product picture using the product picture sorting in primary product in showing the corresponding all products of product category as described demonstration product category in the present embodiment, thereby be automatically embodied as described demonstration product category, set product picture.

Can adopt in addition the product picture of any one product in the corresponding all products of described demonstration product category as the product picture of described demonstration product category.

For example, in described product database, retrieve successively all products that comprise described demonstration product category, and between described product and described demonstration product category, establish the link the product picture using the product picture of any one product in the described product that comprises described demonstration product category as described demonstration product category.

And product database described in step 16 is the product related informations such as name of product, product category and product picture of having stored specific product in prior art.And described product database can be retrieved based on product related informations such as described name of product, product category and product pictures respectively.

Step 17, shows name of product and the product picture of described demonstration product category.Wherein in a viewing area, build a recommendation viewing area, and successively name of product and the product picture of each demonstration product category are shown in described recommendation viewing area according to order from left to right, and the name of product of described demonstration product category is shown in the below of the product picture of described demonstration product category.

So by above-mentioned display mode, the name of product of described demonstration product category and the display location of product picture are outstanding.User is easy to just can click.And adopt the mode Showing Picture, be convenient to user and understand.Be that a so-called figure wins thousand speeches.In addition classify and arrange from left to right with inventory form.Without opening again classification tree, search one by one.Thereby simplified the structure of display interface.

The second embodiment:

In the present embodiment, also by limiting all users in different time intervals are excavated to coupling to the click of keyword search result and buying behavior, thereby lower the error to the search of seasonal product, improve further accuracy and the reliability of Search Results.

Wherein as shown in Figure 2, the classification of the present embodiment qi method that disappears comprises the following steps:

Step 201, reads in a key word.

Step 202, from user's click logs, search first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user.

Wherein said first click ratio and first buy ratio obtain with the first embodiment in identical just repeat no more herein.

Step 203, clicks by first of product category described in each the first relevance parameter that probability and the first purchase probability calculate described product category successively.

Wherein said product category successively through type 2 calculates the first relevance parameter of described product category:

Relevance ₁=w ₁₁* CP ₁(i|q)+w ₁₂* PP ₁(i|q) formula 2

And the first purchase probability weight coefficient w described in it ₁₁be made as 0.7, described first clicks probability right coefficient w ₁₂be made as 0.3.

Step 204, from user's click logs, search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user.

Wherein said the second click ratio and second is bought obtaining in also with the first embodiment of ratio and is identical just repeated no more herein.

Step 205, clicks by second of product category described in each the second relevance parameter that probability and the second purchase probability calculate described product category successively.

Wherein said product category successively through type 3 calculates the second relevance parameter of described product category:

Relevance ₂=w ₂₁* CP ₂(i|q)+w ₂₂* PP ₂(i|q) formula 3

And described the second purchase probability weight coefficient w ₂₁also be 0.7, described second clicks probability right coefficient w ₂₂also be 0.3.

Step 206, the successively relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter.

Wherein said product category successively through type 4 calculates the relative coefficient of described product category:

Relevance_optimized=w ₃* relevance ₁+ w ₄* relevance ₂formula 4

Because the seasonality of some products is more intense.For example the dress ornament in summer and winter just completely different, user's demand also can great changes have taken place.If just only adopt click volume and the sales volume in whole selling time section, have very large deviation when transform season.In the present embodiment by user in certain hour is reduced further to the deviation of seasonal product to the click behavior of the product category that comprises described key word and buying behavior.

And described in step 202 and step 204, the first Preset Time section is less than the second Preset Time section.And the error in order to calm down seasonal product, described the first Preset Time can be made as to 1 week, described the second Preset Time is made as to 1 month.

Step 207, according to the size of relative coefficient, is arranged as a product category sequence according to order from big to small by described product category.

Step 208, in described product category sequence, starts to extract product category that quantity equals described demonstration product number as described demonstration product category from having the product category of maximal correlation property coefficient.

In addition in above-mentioned steps 207 and step 208, can also adopt the mode of searching loop to take out correspondingly relative coefficient.

For example, first from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category, then the numerical value of described demonstration product number is reduced to 1, and whether detect described demonstration product number be zero, if enter step 209, otherwise repeat this step.

Because actual product kind is thousands of, so can comprise a lot of product categories in described product category sequence, but consider the eye-catching object of restriction and the demonstration of actual displayed, so show described in the present embodiment that product number chooses any in 4,5 or 6 too, wherein preferably choose 5.

Step 209, shows each that in a product database product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category successively.

Idiographic flow in described step 209 is identical with the first embodiment, just no longer at length repeats herein.

Step 210, shows name of product and the product picture of described demonstration product category.Wherein in a viewing area, build a recommendation viewing area, and successively name of product and the product picture of each demonstration product category are shown in described recommendation viewing area according to order from left to right, and the name of product of described demonstration product category is shown in the below of the product picture of described demonstration product category.

The 3rd embodiment: a kind of classification qi device that disappears is provided in the present embodiment, and wherein as shown in Figure 3, the classification of the present embodiment qi device that disappears comprises:

One reads in module 1, for reading in a key word.

One retrieval module 2, for search the click ratio of each product category that name of product that user obtains described key search comprises described key word from user's click logs, and buy from user the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record.

One processing module 3, for the relative coefficient that click probability and the purchase probability by product category described in each calculates described product category successively.

One screening module 4, for extracting product category that quantity equals a demonstration product number from described product category as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category.

One picture link module 5, the product picture for the product picture of arbitrary product in a product database shows each included product of product category successively as described demonstration product category.

One display module 6, for showing name of product and the product picture of described demonstration product category.

The relative coefficient that formula 1 in wherein said processing module 3 and retrieval module 2 employing the first embodiment and correlation parameter calculate described product category.

Described screening module 4 comprises an order module 41, for by described product category according to the size of relative coefficient, according to order from big to small, be arranged as a product category sequence.And an extraction module 42, in described product category sequence, from thering is the product category of maximal correlation property coefficient, start to extract product category that quantity equals described demonstration product number as described demonstration product category.

And described picture link module 5 comprises that one for retrieving all products that comprise described demonstration product category at described product database successively, and between described product and described demonstration product category, establish the link the module using the product picture of any one product in the described product that comprises described demonstration product category as the product picture of described demonstration product category.

Described display module 6 is for build a recommendation viewing area in a viewing area, and is shown in described recommendation viewing area according to name of product and the product picture that order from left to right shows product category by each successively.And in described display module 6, also comprise that one for being shown in the name of product of described demonstration product category the module of below of the product picture of described demonstration product category.

For convenience of description, the qi device that in the present embodiment, described classification disappeared is divided into various modules according to function and describes respectively, so when implementing the present embodiment, the function of each module can be realized in same or a plurality of software and/or hardware.

The 4th embodiment

As shown in Figure 3, the disappear difference of qi device and the 3rd embodiment of the classification of the present embodiment is: described in the present embodiment, screen module 4 and comprise:

One searches module 43, for from described product category, extract there is maximal correlation property coefficient product category as described demonstration product category.And a counting module 44, whether for the numerical value of described demonstration product number is reduced to 1, and to detect described demonstration product number be zero, if described picture link module executable operations, otherwise search module described in again calling.

In addition the link module of picture described in the present embodiment 5 is for retrieving all products that comprise described demonstration product category at described product database successively, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

The 5th embodiment:

A kind of classification qi device that disappears is provided in the present embodiment, and wherein as shown in Figure 5, the classification of the present embodiment qi device that disappears comprises:

One reads in module 1, for reading in a key word.

One first retrieval module 7, for search first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word from user's click logs, click ratio, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user;

One first processing module 8, for clicking by first of product category described in each the first relevance parameter that probability and the first purchase probability calculate described product category successively;

One second retrieval module 9, for search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word from user's click logs, click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user;

One second processing module 10, for clicking by second of product category described in each the second relevance parameter that probability and the second purchase probability calculate described product category successively;

One the 3rd processing module 11, for the relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter successively;

The relative coefficient that all adopts formula 2, formula 3 and formula 4 in the second embodiment and correlation parameter to calculate described product category in wherein said the first retrieval module 7, the first processing module 8, the second retrieval module 9, the second processing module 10 and the 3rd processing module 11.

The 6th embodiment:

As shown in Figure 6, the disappear difference of qi device and the 5th embodiment of the classification of the present embodiment is: described in the present embodiment, screen module 4 and comprise:

The 7th embodiment:

Classification in the present embodiment qi system that disappears realizes disappear qi method and the mode of device by essential general hardware platform thereof of above-mentioned classification.

Wherein as shown in Figure 7, the described classification qi system that disappears comprises:

One storer 100, wherein stores a product database.

One input media 101, for reading in a key word.

One server 102, described server 102 comprises that user's click logs and a user buy daily record, wherein said server 102 is searched the click ratio of each product category that name of product that user obtains described key search comprises described key word from user's click logs, and buy the purchase ratio of searching each product category that name of product that user obtains described key search comprises described key word daily record from user, and the relative coefficient that click probability and the purchase probability by product category described in each calculates described product category successively.

One processor 103, for extracting product category that quantity equals a demonstration product number from described product category as showing product category, the relative coefficient of described demonstration product category is all more than or equal to the relative coefficient of remaining product category in described product category, and in the product database of described storer, successively each is shown to the product picture of arbitrary product in the included product of product category is as the product picture of described demonstration product category.

One display 104, for showing name of product and the product picture of described demonstration product category.

Wherein said data processing method and product database ground can be with above-described embodiment identical, be just no longer described in detail herein.

The 8th embodiment:

Classification in the present embodiment also above-mentioned classification is disappeared qi method and the mode of device by essential general hardware platform thereof of qi system that disappear realizes.

Wherein described in the present embodiment, the disappear difference of qi system and the 7th embodiment of classification is:

Server 102 described in the server 102 of the present embodiment comprises that user's click logs and a user buy daily record, wherein said server 102 is searched first of each product category that user obtains described key search in one first Preset Time section name of product comprises described key word and is clicked ratio from user's click logs, and buy and daily record, search first of each product category that user obtains described key search in described the first Preset Time section name of product comprises described key word and buy ratio from user, and from user's click logs, search second of each product category that user obtains described key search in one second Preset Time section name of product comprises described key word and click ratio, and buy and daily record, search second of each product category that user obtains described key search in described the second Preset Time section name of product comprises described key word and buy ratio from user, and described server 102 also clicks by first of product category described in each the first relevance parameter that probability and the first purchase probability calculate described product category successively, and by second of product category described in each, click the second relevance parameter that probability and the second purchase probability calculate described product category successively, then the relative coefficient of product category described in the first relevance parameter by product category described in each and the second correlativity calculation of parameter successively.

As seen through the above description of the embodiments, those skilled in the art can be well understood to the mode that the application can add essential general hardware platform by software and realizes.Understanding based on such, the part that the application's technical scheme contributes to prior art in essence in other words can embody with the form of software product, this computer software product can be stored in storage medium, as ROM/RAM, magnetic disc, CD etc., comprise that some instructions are with so that a computer equipment (can be personal computer, server, or the network equipment etc.) carry out the method described in some part of each embodiment of the application or embodiment.

Each embodiment in this instructions all adopts the mode of going forward one by one to describe, between each embodiment identical similar part mutually referring to, each embodiment stresses is the difference with other embodiment.Especially, for system embodiment, because it is substantially similar in appearance to embodiment of the method, so description is fairly simple, relevant part is referring to the part explanation of embodiment of the method.

The application can be used in numerous general or special purpose computingasystem environment or configuration.For example: personal computer, server computer, handheld device or portable set, plate equipment, multicomputer system, the system based on microprocessor, set top box, programmable consumer-elcetronics devices, network PC, small-size computer, mainframe computer, comprise distributed computing environment of above any system or equipment etc.

Although more than described the specific embodiment of the present invention, it will be understood by those of skill in the art that these only illustrate, protection scope of the present invention is limited by appended claims.Those skilled in the art is not deviating under the prerequisite of principle of the present invention and essence, can make various changes or modifications to these embodiments, but these changes and modification all fall into protection scope of the present invention.

Claims

1. the classification qi method that disappears, is characterized in that, the described classification qi method that disappears comprises the following steps:

S ₁, read in a key word;

2. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₃for:

Relevance=w ₁* CP (i|q)+w ₂* PP (i|q) formula 1

3. the classification as claimed in claim 2 qi method that disappears, is characterized in that described purchase probability weight coefficient w ₁be 0.7, described click probability right coefficient w ₂be 0.3.

4. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₄for:

5. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₄for:

6. the qi method that disappears of the classification as described in claim 4 or 5, is characterized in that, described demonstration product number is 4,5 or 6.

7. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₅for:

8. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₅for:

9. the classification as claimed in claim 1 qi method that disappears, is characterized in that described step S ₆for:

10. the classification as claimed in claim 9 qi method that disappears, is characterized in that described step S ₆in further comprising the steps of:

11. 1 kinds of classification qi method that disappears, is characterized in that, the described classification qi method that disappears comprises the following steps:

S ₁₀₁, read in a key word;

The 12. classification as claimed in claim 11 qi method that disappears, is characterized in that, described the first Preset Time section is less than the second Preset Time section.

The 13. classification as claimed in claim 12 qi method that disappears, is characterized in that, described the first Preset Time is 1 week, and described the second Preset Time is 1 month.

The 14. classification as claimed in claim 12 qi method that disappears, is characterized in that described step S ₁₀₃for:

Relevance ₁=w ₁₁* CP ₁(i|q)+w ₁₂* PP ₁(i|q) formula 2

The 15. classification as claimed in claim 14 qi method that disappears, is characterized in that described the first purchase probability weight coefficient w ₁₁be 0.7, described first clicks probability right coefficient w ₁₂be 0.3.

The 16. classification as claimed in claim 12 qi method that disappears, is characterized in that described step S ₁₀₅for:

Relevance ₂=w ₂₁* CP ₂(i|q)+w ₂₂* PP ₂(i|q) formula 3

The 17. classification as claimed in claim 16 qi method that disappears, is characterized in that described the second purchase probability weight coefficient w ₂₁be 0.7, described second clicks probability right coefficient w ₂₂be 0.3.

18. classification as described in claim 14 and the 16 qi method that disappears, is characterized in that described step S ₁₀₆for:

Relevance_optimized=w ₃* relevance ₁+ w ₄* relevance ₂formula 4

The 19. classification as claimed in claim 11 qi method that disappears, is characterized in that described step S ₁₀₇for:

The 20. classification as claimed in claim 11 qi method that disappears, is characterized in that described step S ₁₀₇for:

21. classification as described in the claim 19 or 20 qi method that disappears, is characterized in that, described demonstration product number is 4,5 or 6.

The 22. classification as claimed in claim 11 qi method that disappears, is characterized in that described step S ₁₀₈for:

The 23. classification as claimed in claim 11 qi method that disappears, is characterized in that described step S ₁₀₈for:

The 24. classification as claimed in claim 11 qi method that disappears, is characterized in that described step S ₁₀₉for:

The 25. classification as claimed in claim 24 qi method that disappears, is characterized in that described step S ₁₀₉in further comprising the steps of:

26. 1 kinds of classification qi device that disappears, is characterized in that, the described classification qi device that disappears comprises:

One reads in module, for reading in a key word;

The 27. classification as claimed in claim 26 qi device that disappears, is characterized in that, calculates the relative coefficient of described product category in product category described in described processing module successively through type 5:

Relevance=w ₁* CP (i|q)+w ₂* PP (i|q) formula 5

The 28. classification as claimed in claim 27 qi device that disappears, is characterized in that described purchase probability weight coefficient w ₁be 0.7, described click probability right coefficient w ₂be 0.3.

The 29. classification as claimed in claim 26 qi device that disappears, is characterized in that, described screening module comprises:

The 30. classification as claimed in claim 26 qi device that disappears, is characterized in that, described screening module comprises:

31. classification as described in the claim 29 or 30 qi device that disappears, is characterized in that, described demonstration product number is 4,5 or 6.

The 32. classification as claimed in claim 26 qi device that disappears, it is characterized in that, described picture link module comprises that one for retrieving all products that comprise described demonstration product category at described product database successively, and between described product and described demonstration product category, establish the link the module using the product picture of any one product in the described product that comprises described demonstration product category as the product picture of described demonstration product category.

The 33. classification as claimed in claim 26 qi device that disappears, it is characterized in that, described picture link module is for retrieving all products that comprise described demonstration product category at described product database successively, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

The 34. classification as claimed in claim 26 qi device that disappears, it is characterized in that, described display module is used for building a recommendation viewing area in a viewing area, and successively name of product and the product picture of each demonstration product category is shown in described recommendation viewing area according to order from left to right.

The 35. classification as claimed in claim 34 qi device that disappears, is characterized in that, also comprises that one for being shown in the name of product of described demonstration product category the module of below of the product picture of described demonstration product category in described display module.

36. 1 kinds of classification qi device that disappears, is characterized in that, the described classification qi device that disappears comprises:

One reads in module, for reading in a key word;

The 37. classification as claimed in claim 36 qi device that disappears, is characterized in that, described the first Preset Time section is less than the second Preset Time section.

The 38. classification as claimed in claim 37 qi device that disappears, is characterized in that, described the first Preset Time is 1 week, and described the second Preset Time is 1 month.

The 39. classification as claimed in claim 36 qi device that disappears, is characterized in that, product category described in described the first processing module successively through type 6 calculates the first relevance parameter of described product category:

Relevance ₁=w ₁₁* CP ₁(i|q)+w ₁₂* PP ₁(i|q) formula 6

The 40. classification as claimed in claim 39 qi device that disappears, is characterized in that described the first purchase probability weight coefficient w ₁₁be 0.7, described first clicks probability right coefficient w ₁₂be 0.3.

The 41. classification as claimed in claim 36 qi device that disappears, is characterized in that, product category described in described the second processing module successively through type 7 calculates the second relevance parameter of described product category:

Relevance ₂=w ₂₁* CP ₂(i|q)+w ₂₂* PP ₂(i|q) formula 7

The 42. classification as claimed in claim 41 qi device that disappears, is characterized in that described the second purchase probability weight coefficient w ₂₁be 0.7, described second clicks probability right coefficient w ₂₂be 0.3.

The 43. classification as claimed in claim 36 qi device that disappears, is characterized in that, product category described in described the 3rd processing module successively through type 8 calculates the relative coefficient of described product category:

Relevance_optimized=w ₃* relevance ₁+ w ₄* relevance ₂formula 8

The 44. classification as claimed in claim 36 qi device that disappears, is characterized in that, described screening module comprises:

The 45. classification as claimed in claim 36 qi device that disappears, is characterized in that, described screening module comprises:

46. classification as described in the claim 44 or 45 qi device that disappears, is characterized in that, described demonstration product number is 4,5 or 6.

The 47. classification as claimed in claim 36 qi device that disappears, it is characterized in that, described picture link module comprises that one for retrieving all products that comprise described demonstration product category at described product database successively, and between described product and described demonstration product category, establish the link the module using the product picture of any one product in the described product that comprises described demonstration product category as the product picture of described demonstration product category.

The 48. classification as claimed in claim 36 qi device that disappears, it is characterized in that, described picture link module is for retrieving all products that comprise described demonstration product category at described product database successively, and comprise that by described the product sequence of described demonstration product category is a sequence, to between described sequence and described demonstration product category, establish the link, and the product picture using the product picture of first product in described sequence as described demonstration product category.

The 49. classification as claimed in claim 36 qi device that disappears, it is characterized in that, described display module is used for building a recommendation viewing area in a viewing area, and successively name of product and the product picture of each demonstration product category is shown in described recommendation viewing area according to order from left to right.

The 50. classification as claimed in claim 49 qi device that disappears, is characterized in that, also comprises that one for being shown in the name of product of described demonstration product category the module of below of the product picture of described demonstration product category in described display module.

51. 1 kinds of classification qi system that disappears, is characterized in that, the described classification qi system that disappears comprises:

One storer, wherein stores a product database;

One input media, for reading in a key word;

52. 1 kinds of classification qi system that disappears, is characterized in that, the described classification qi system that disappears comprises:

One storer, wherein stores a product database;

One input media, for reading in a key word;

The 53. classification as claimed in claim 52 qi system that disappears, is characterized in that, described the first Preset Time section is less than the second Preset Time section.