Method, system, and computer program product for recommending online products

Patent metadata

Field	Value
Publication number	US-9984402-B2
Application number	US-201414461139-A
Country	US
Kind code	B2
Filing date	Aug 15, 2014
Priority date	Aug 26, 2013
Publication date	May 29, 2018
Grant date	May 29, 2018

Abstract

Embodiments of the present application relate to a method for recommending online products, a system for recommending online products, and a computer program product for recommending online products. A method for recommending online products is provided. The method includes specifying a main product zone of a query product image, dividing the main product zone into a plurality of local zones, extracting color features from each local zone, looking up candidate recommended product images sharing common characteristics with a query product image based on the color features of each local zone, matching, among the found candidate recommended product images, product images that are similar in terms of color matching to the query product image, and regarding the matched product images as recommended product images.

First claim

What is claimed is:

1. A method for recommending online products, comprising:
specifying, by a processor, a main product zone of a query product image;
removing, by the processor, an image background from the main product zone to obtain a modified main product zone;
calculating, by the processor, a center of gravity of the modified main product zone;
dividing, by the processor, the modified main product zone into a plurality of local zones based on the calculated center of gravity;
extracting, by the processor, color features from a local zone, wherein the extracting is performed by dividing the local zone into a plurality of portions using a hue, saturation, and brightness space;
looking up, by the processor, candidate recommended product images sharing common characteristics with the query product image based on the color features of the local zone;
matching, among the found candidate recommended product images, product images that are similar in terms of color matching to the query product image;
providing the matched product images as recommended product images to a client for display; and
establishing, by the processor an inverted index for the product images based on numbers of words in the product images, wherein:
the main product zone is comprised of high-dimensional vectors, a high-dimensional vector representing a vector corresponding to a distribution of a plurality of dimensions, a dimension corresponding to a color.

2. The method for recommending online products as described in claim 1 , wherein the looking up of the candidate recommended product images sharing the common characteristics with the query product image comprises:
accessing entries of the inverted index corresponding to word frequencies that occur within a query product image P to calculate a similarity between the query product image P and a recommendable product image Q based on a cosine of an angle in a high-dimensional space formed by two high-dimensional vectors composed of the word frequencies of the query product image P and the recommendable product image Q; and
selecting a top-ranked preset quantity M recommendable product images from the calculated similarity as candidate recommended product images, wherein M is an integer greater than one.

3. The method for recommending online products as described in claim 2 , wherein the calculating of the similarities between the query product image P and the recommendable product image Q comprises:

sim
⁡

(

P
,
Q

)

=

∑

i
=
1

w

⁢

⁢

p
i

⁢

q
i


P


⁢


Q


,

wherein w represents a quantity of high-dimensional vectors contained in each product image, w=r×m, r represents a quantity of local zones into which each image has been divided, m represents a quantity of colors contained in the local zone, ∥ represents a modulus operator,


P


=

∑

i
=
1

w

⁢

⁢

p
i
2

,


Q


=

∑

i
=
1

w

⁢

⁢

q
i
2

,

represents a word frequency corresponding to an i th high-dimensional vector element within the query product image P, q i represents a word frequency corresponding to the i th high-dimensional vector element within the recommendable product image Q, and a background zone word frequency is zero.

4. The method for recommending online products as described in claim 3 , wherein the matching of the product images that are similar in terms of color matching to the query product image comprises:
performing matching calculations on one of the candidate recommended product images at a time;
reducing differences between the one candidate recommended product image and the query product image with respect to these color intervals based on dominant hues shared by two images that are matched;
strengthening differences between the one candidate recommended product image and the query product image with respect to these color intervals based on details or different color parts of the two images that are matched to obtain a color matching similarity colorsim (P,Q) for each candidate recommended product image Q and the query product image P; and
performing one of the following:
1) acquiring a preset quantity N product images top-ranked by similarity from among the candidate recommended product images and regarding the preset quantity N product images as recommended product images similar in terms of color matching to the query product image, wherein N is an integer greater than one; or
2) selecting, from among the candidate recommended product images, product images whose similarity to the query product image is greater than a preset similarity threshold value and regarding the product images as recommended product images.

5. The method for recommending online products as described in claim 4 , wherein the obtaining of the color matching similarity colorsim (P,Q) of each candidate recommended product image Q and the query product image P is:

colorsim
⁡

Assignees

• Alibaba Group Holding Ltd

Inventors

• Cao Yang

Classifications

• G06Q30/0631 Primary

  Recommending goods or services · CPC title

• G06Q30/0222

  During e-commerce, i.e. online transactions · CPC title

• G06F16/5838

  using colour · CPC title

• G06Q30/0623

  by investigating goods or services · CPC title

• G06Q30/0629

  by pre-processing results, e.g. ranking or ordering results · CPC title