Dynamic taxonomy generation with demand-based product groups

Patent metadata

Field	Value
Publication number	US-9613371-B2
Application number	US-201414475325-A
Country	US
Kind code	B2
Filing date	Sep 2, 2014
Priority date	Sep 2, 2014
Publication date	Apr 4, 2017
Grant date	Apr 4, 2017

Abstract

A method of extending an existing product taxonomy for an inventory of products. The existing product taxonomy can include pages having a hierarchical tree structure. The method can include determining a first set of keywords. The first set of keywords can be a prediction of keywords searched above a predetermined threshold during a predetermined time period of one or more external search engines. The method also can include generating dynamic product groups based on the first set of keywords. The method further can include generating a relation graph of the dynamic product groups. The relation graph can include relation links between the dynamic product groups. The method also can include linking the dynamic product groups to pages in the existing product taxonomy to generate an extended product taxonomy. The extended product taxonomy can include the existing product taxonomy, the dynamic product groups, and the relation links of the relation graph between the dynamic product groups. The method further can include presenting to a user information about the dynamic product groups. Other embodiments are provided.

First claim

What is claimed is:

1. A method of extending an existing product taxonomy for an inventory of products, the existing product taxonomy comprising pages having a hierarchical tree structure, the method being implemented via execution of computer instructions configured to run at one or more processing modules and configured to be stored at one or more non-transitory memory storage modules, the method comprising:
determining a first set of keywords, the first set of keywords being a prediction of keywords searched above a predetermined threshold during a predetermined time period of one or more external search engines, wherein the determining the first set of keywords comprises using a linear-regression time-series machine-learning model to predict the keywords searched above the predetermined threshold during the predetermined time period of the one or more external search engines;
generating dynamic product groups based on the first set of keywords comprising:
generating the dynamic product groups from keyword clusters comprising:
searching the inventory of products using the keyword clusters to generate relevant item clusters;
determining a category path for each of the relevant item clusters comprising:
building a first tree for each of the relevant item clusters, the first tree comprising first node corresponding to nodes in the hierarchical tree structure of the existing product taxonomy, the first node comprising relevant items of the relevant item clusters;
pruning the first tree to remove nodes having relevant items with lower relevance scores;
suppressing one or more subtrees of the first tree to create one or more super-nodes, the one or more subtrees having child nodes having relevant items with relevance scores that are approximately uniformly distributed across the child nodes;
determining a first leaf node of the first tree having a highest relevance; and
selecting the category path based on a path of the first tree to the first leaf node; and

applying one or more facet selection filters to the relevant item clusters to generate the dynamic product groups;

generating a relation graph of the dynamic product groups, the relation graph comprising relation links between the dynamic product groups;
linking the dynamic product groups to pages in the existing product taxonomy to generate an extended product taxonomy, the extended product taxonomy comprising the existing product taxonomy, the dynamic product groups, and the relation links of the relation graph between the dynamic product groups; and
presenting to a user information about the dynamic product groups.

2. The method of claim 1 , wherein determining the first set of keywords comprises:
determining the first set of keywords based on at least one of internal search volumes or search engine marketing volumes, and based on a top query list for the one or more external search engines.

3. The method of claim 1 , wherein generating the dynamic product groups comprises;
clustering the first set of keywords based on similarity of intent to generate the keyword clusters.

4. The method of claim 1 , wherein generating the relation graph of the dynamic product groups comprises:
generating the relation graph of the dynamic product groups such that the relation graph comprises a parent-child relation graph, wherein the relation links between the dynamic product groups are based on title suffix matching.

5. The method of claim 1 , wherein generating the relation graph of the dynamic product groups comprises:
generating the relation graph of the dynamic product groups such that the relation graph comprises a semantic relation graph, wherein the relation links between the dynamic product groups are based on a similarity of items in the dynamic product groups.

6. The method of claim 5 , wherein;
the similarity of items in the dynamic product groups is measured by a Jaccard index.

7. The method of claim 1 , wherein generating the relation graph of the dynamic product groups comprises:
generating the relation graph of the dynamic product groups such that the relation graph comprises a co-visit relation graph, wherein the relation links between the dynamic product groups are based on a co-visit history of the dynamic product groups.

8. The method of claim 1 , wherein generating the relation graph of the dynamic product groups comprises:
generating the relation graph of the dynamic product groups such that the relation graph comprises a superimposition of at least two of: a parent-child relation graph, a semantic relation graph, or a co-visit relation graph.

9. The method of claim 1 , wherein linking the dynamic product groups to pages in the existing product taxonomy to generate the extended product taxonomy comprises:
adding hypedlinks to the pages in the existing product taxonomy to link to the dynamic product groups based on category paths for the dynamic product groups.

10. The method of claim 1 , wherein presenting to the user information about the dynamic product groups comprises:
presenting to the user a taxonomy path of the extended product taxonomy for one of the dynamic product groups.

11. The method of claim 1 , wherein presenting to the user information about the dynamic product groups comprises:
presenting to the user related links for one of the dynamic product groups based on the relation links of the relation graph corresponding to the one of the dynamic product groups.

12. The method of claim 1 , wherein presenting to the user information about the dynamic product groups comprises:
presenting to the user a mashup of dynamic product groups based on category paths of the dynamic product groups end the relation links of the relation graph.

13. A system for extending an existing product taxonomy for an inventory of products, the existing product taxonomy comprising pages having a hierarchical tree structure, the system comprising:
one or more processing modules; and
one or more non-transitory memory storage modules storing computing instructions configured to run on the one or more processing modules and perform the acts of;
determining a first set of keywords, the first set of keywords being a prediction of keywords searched above a predetermined threshold during a predetermined time period of one or more external search engines;
generating dynamic product groups based on the first set of keywords comprising:
generate the dynamic product groups from keyword clusters comprising:
searching the inventory of products using the keyword clusters to generate relevant item clusters;
determining a category path for each of the relevant item clusters comprising:
building a first tree for each of the relevant item clusters, the first tree comprising first node corresponding to nodes in the hierarchical tree structure of the existing product taxonomy, the first node comprising relevant items of the relevant item clusters;
pruning the first tree to remove nodes having relevant items with lower relevance scores;
suppressing one or more subtrees of the first tree to create one or more super-nodes, the one or more subtrees having child nodes having relevant items with relevance scores that are approximately uniformly distributed across the child nodes;
determining a first leaf node of the first tree having a highest relevance; and
selecting the category path based on a path of the first tree to the first leaf node; and

applying one or more facet selection filters to the relevant item clusters to generate the dynamic product groups;

generating a relation graph of the dynamic product groups, the relation graph comprising:
relation links between the dynam

Assignees

• Wal Mart Stores Inc

Inventors

• Subramanya Shankara B
• Yuan Guoxun
• Mathihalli Madhusudan
• Tal Ron

Classifications

• G06F17/30958

  Physics · mapped topic

• G06Q10/087

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title

• G06F17/30734

  Physics · mapped topic

• G06F17/30873

  Physics · mapped topic

• G06Q30/0623 Primary

  by investigating goods or services · CPC title