Vertical Search-Based Query Method, System and Apparatus

• US 20130054569A1
• Filed: 04/26/2011
• Published: 02/28/2013
• Est. Priority Date: 04/30/2010
• Status: Active Grant

• Current Assignee: Alibaba Group Holding Ltd.
• Original Assignee: Alibaba Group Holding Ltd.
• Inventors: Ning, Wei, Nie, Xuping, Xiao, He, Fu, Qile, Mo, Yi
• Granted Patent: US 8,661,027 B2
• Time in Patent Office: Days
• US Class Current: 707/722
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/248 Presentation of query results
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

Various embodiments of a method, system, and apparatus related to query based on vertical search are disclosed. In one aspect, a method of query based on vertical search receives a user query. The method obtains a first category model from a category model warehouse based on the user query to generate a first query result. The first category model includes one or more commodity categories that correspond to one or more keywords in the user query. The method also obtains one or more commodity categories corresponding to the user query from a commodity warehouse to generate a second query result. The method further generates a final query result by combining the first query result and the second query result.

Claims

1. A method of query based on a vertical search, the method comprising:

receiving a user query;

obtaining a first category model from a category model warehouse based on the user query to generate a first query result, the first category model including one or more commodity categories that correspond to one or more keywords in the user query;

obtaining one or more commodity categories corresponding to the user query from a commodity warehouse to generate a second query result; and

generating a final query result by combining the first query result and the second query result.

2. A method as recited in claim 1, further comprising:

transmitting the final query result to a user;

generating a log based on the user query and the user'"'"'s clicking operations in response to the final query result;

generating a second category model from statistical analysis of data in the log; and

updating the category model warehouse with the second category model.

3. A method as recited in claim 1, wherein the first category model comprises one or more attribute categories corresponding to the one or more commodity categories, and wherein generating the second query result comprises obtaining one or more commodity categories and associated attribute categories that match the user query.

4. A method as recited in claim 1, wherein, when the user query consists of one or more keywords inputted by a user, the obtaining the first category model from the category model warehouse based on the user query to generate the first query result comprises:

determining whether a keyword corresponding to the first category model matches any of the one or more keywords in the user query;

in an event that the keyword corresponding to the first category model matches any of the one or more keywords in the user query;

obtaining a third category model that matches a keyword of the one or more keywords in the user query that matches in the keyword corresponding to the first category model; and

generating the first query result using the third category model, one or more corresponding weights, one or more attribute categories of one or more direct attributes, and one or more weights corresponding to the one or more attribute categories;

in an event that keywords corresponding to the first category model do not include any of the one or more keywords in the user query;

revising the one or more keywords in the user query and repeating the determining until a fourth category model that matches any of the revised one or more keywords in the user query is obtained; and

generating the first query result using the third category model or the fourth category model, one or more corresponding weights, one or more attribute categories of one or more direct attributes, and one or more weights corresponding to the one or more attribute categories.

5. A method as recited in claim 1, wherein, when the user query comprises one or more keywords inputted by a user and one or more commodity categories selected by the user, the obtaining the first category model from the category model warehouse based on the user query to generate the first query result comprises:

determining whether keywords corresponding to the first category model include any of the one or more keywords in the user query;

in an event that a keyword corresponding to the first category model matches any of the one or more keywords in the user query;

obtaining a third category model that matches a keyword of the one or more keywords in the user query that matches in the keyword corresponding to the first category model; and

generating the first query result using the third category model, one or more corresponding weights, one or more attribute categories of one or more direct attributes, and one or more weights corresponding to the one or more attribute categories;

in an event that the keyword corresponding to the first category model does not match any of the one or more keywords in the user query;

revising the one or more keywords in the user query and repeating the determining until a fourth category model that matches any of the revised one or more keywords in the user query is obtained; and

obtaining a fifth category model that matches the one or more commodity categories selected by the user in the user query from the third category model or the fourth category model; and

generating the first query result using commodity categories in the fifth category model and corresponding weights thereof

6. A method as recited in claim 1, wherein the generating the final query result by combining the first query result and the second query result comprises:

obtaining a first combined result comprising commodity categories that are in both the first query result and the second query result, the first combined result further comprising weights corresponding to the commodity categories in the first combined result and being a combination of those weights from the first query result and from the second query result;

obtaining a second combined result comprising commodity categories and corresponding weights in the second query result;

increasing the weights in the first combined result to render weights corresponding to the commodity categories in the first combined result to be respectively higher than the weights corresponding to the commodity categories in the second combined result; and

arranging the commodity categories in the first combined result according to the corresponding weights in a descending order to provide the first combined result to a user.

7. A method as recited in claim 3, wherein the generating the final query result by combining the first query result and the second query result comprises:

obtaining a first combined result comprising commodity categories and associated attribute categories that are in both the first query result and the second query result, the first combined result further comprising weights corresponding to the commodity categories and the associated attribute categories in the first combined result and being a combination of those weights from the first query result and the second query result;

obtaining a second combined result comprising commodity categories and associated attribute categories in the second query result, the second combined result further comprising weights corresponding to the commodity categories and the associated attribute categories in the second query result;

increasing the weights in the first combined result to render weights corresponding to the commodity categories and attribute categories corresponding to the commodity categories in the first combined result to be respectively higher than the weights corresponding to the commodity categories and attribute categories corresponding to the commodity categories in the second combined result; and

arranging the commodity categories and the associated attribute categories in the first combined result according to the corresponding weights in a descending order to provide the first combined result to a user.

8. A method as recited in claim 2, wherein the generating the log based on the user query and the user'"'"'s clicking operations in response to the final query result comprises:

obtaining data on the user'"'"'s clicking operations with respect to commodity categories, attribute categories corresponding to the commodity categories, and commodities in response to the final query result;

generating the log based on the data on the user'"'"'s clicking operations, the log comprising the user query and corresponding clicking information, the clicking information comprising a commodity category and an attribute category of a commodity clicked on by the user, a commodity category clicked on by the user, and a attribute category clicked on by the user; and

storing the generated log.

9. A method as recited in claim 2, wherein the generating the second category model from statistical analysis of data in the log comprises:

obtaining a statistical analysis result based on statistical analysis of the user query and corresponding clicking information in the log, the statistical analysis result comprising commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query, the weights corresponding to the commodity categories and the associated attribute categories comprising number of clicks on the commodity categories and the associated attribute categories, clicking probability related to the commodity categories and the associated attribute categories, or a combination thereof; and

arranging the statistical analysis result as a category tree to generate the second category model.

10. A method as recited in claim 9, wherein the arranging the statistical analysis result as a category tree to generate the second category model comprises:

determining whether the commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query reach one or more weight thresholds; and

in an event that the one or more weight thresholds are reached, generating the category model based on the commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query.

11. A query system based on a vertical search, comprising:

a query server that;

receives a user query;

obtains a first category model from a category model warehouse based on the user query to generate a first query result, the first category model including one or more commodity categories that correspond to one or more keywords in the user query;

obtains one or more commodity categories corresponding to the user query from a commodity warehouse to generate a second query result; and

generates a final query result by combining the first query result and the second query result;

a log server that generates a log based on the user query and user'"'"'s clicking operations in response to the final query result; and

a modeling server that generates a second category model from statistical analysis of data in the log.

12. A query system as recited in claim 11, wherein the modeling server transmits the second category model to the query server, wherein the query server provides the final query result to a user, and updates the category model warehouse with the second category model.

13. A query system as recited in claim 11, wherein the first category model comprises one or more attribute categories corresponding to the one or more commodity categories, and wherein the query server obtains one or more commodity categories and attribute categories corresponding to the user query from the commodity warehouse.

14. A query system as recited in claim 11, wherein the query server obtains data on the user'"'"'s clicking operations with respect to commodity categories, attribute categories corresponding to the commodity categories, and commodities in response to the final query result, wherein the log server generates the log based on the data on the user'"'"'s clicking operations, the log comprising the user query and corresponding clicking information, the clicking information comprising a commodity category and an attribute category of a commodity clicked on by the user, a commodity category clicked on by the user, and a attribute category clicked on by the user, and wherein the log server further stores the generated log.

15. A query system as recited in claim 11, wherein the modeling server further:

obtains a statistical analysis result based on statistical analysis of the user query and corresponding clicking information in the log, the statistical analysis result comprising commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query, the weights corresponding to the commodity categories and the associated attribute categories comprising number of clicks on the commodity categories and the associated attribute categories, clicking probability related to the commodity categories and the associated attribute categories, or both; and

arranges the statistical analysis result as a category tree to generate the second category model.

16. A query system as recited in claim 15, wherein the modeling server further:

determines whether the commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query reach one or more weight thresholds; and

in an event that the one or more weight thresholds are reached, generates the category model based on the commodity categories, associated attribute categories, and weights corresponding to the commodity categories and the associated attribute categories that correspond to the user query.

17. A query server, comprising:

an acquisition module that obtains a user query;

a query module that, based on the user query, retrieves a category model matching the user query from a category model warehouse, generates a first query result based on the category model, and searches for commodity categories that match the user query in a commodity warehouse to generate a second search result, the category model comprising one or more commodity categories corresponding to one or more keywords in the user query; and

a combination module that combines the first query result and the second query result to generate a final query result.

18. A query server as recited in claim 17, further comprising a transmission module that:

transmits the final query result to a user;

generates a log based on the user query and the user'"'"'s clicking operations in response to the final query result;

generates a second category model from statistical analysis of data in the log; and

updates the category model warehouse with the second category model.

19. A server as recited in claim 17, wherein the first category model comprises one or more attribute categories corresponding to the one or more commodity categories, and wherein the query module obtains one or more commodity categories and associated attribute categories that match the user query.