Method and apparatus for sorting related searches

Patent metadata

Field	Value
Publication number	US-10331685-B2
Application number	US-201715625936-A
Country	US
Kind code	B2
Filing date	Jun 16, 2017
Priority date	Dec 29, 2016
Publication date	Jun 25, 2019
Grant date	Jun 25, 2019

Abstract

The present disclosure provides a method and apparatus for sorting related searches RS, wherein the method comprises: obtaining a query and obtaining RSes corresponding to the query; determining estimated click rates of the RSes respectively according to a click estimating model obtained from pre-training; sorting the RSes in a descending order according to the estimated click rates. The solutions of the present disclosure can improve the accuracy of the sorting results.

First claim

What is claimed is:

1. A method of sorting related searches on Internet, wherein the method comprises:
obtaining, by a computer via the Internet, a query and obtaining related searches corresponding to the query on the Internet;
determining, by the computer, estimated click rates of the related searches respectively according to a click estimating model obtained from pre-training, wherein the determining includes considering a user behavior feature as input of the click estimating model to obtain an estimated click rate of a related search output by the click estimating model;
sending, by the computer, the estimated click rates of the related searches from a first processing subunit to a second processing subunit to respectively determine guiding capability scores of the related searches according to the estimated click rates and guiding capability features of the related searches;
sorting, by the computer, the related searches based on the estimated click rates that are determined according to the click estimating model, wherein the sorting of the related searches based on the estimated click rates comprises:
sorting, by a sorting subunit of the computer, all related searches in a descending order according to the guiding capability scores, wherein the respective determination of the guiding capability scores of the related searches according to the estimated click rates and the guiding capability features of the related searches comprises:
during user's searching on the computer, performing, by the second processing subunit of the computer, following operation for each related search of the related searches:
respectively obtaining click times of the related searches displayed on the computer within a recent predetermined time length after the user clicks the related search, adding up the click times, dividing a sum by times of displaying the related searches within the recent predetermined time length after the related search is clicked, to obtain a guiding capability feature of the related search;
obtaining a click rate of the related search within the recent predetermined time length; and
calculating a product of the guiding capability feature of the related search and the click rate of the related search within the recent predetermined time length, and adding up the product with an estimated click rate of the related search after having gone through normalization processing to obtain a guiding capability score of the related search; and

after sorting all the related searches in the descending order according to the guiding capability scores and the estimated click rates to improve accuracy of the sorting of all the related searches, displaying, by the computer, top N related searches in a search result page, and recommending the displayed top N related searches to a user, the N is a positive integer.

2. The method according to claim 1 , wherein the determining estimated click rates of the related searches respectively according to a click estimating model obtained from pre-training comprises:
for each related search, obtaining a user behavior feature of the related search respectively.

3. The method according to claim 2 , wherein the user behavior feature comprises: co-occurrence feature, quality control feature, click feature and semantic feature.

4. The method according to claim 1 , wherein the respectively determining quality scores of related searches according to quality features of related searches comprises:
as for each related search, determining quality score of the related search in a fitting manner respectively according to the quality control feature, guiding capability feature and estimated click rate of the related search.

5. A device for sorting related searches on Internet, comprising:
at least one processor of a computer; and
a memory storing instructions, which when executed by the at least one processor, cause the at least one processor to perform operations, the operations comprising:
obtaining, via the Internet, a query and obtaining related searches corresponding to the query on the Internet;
determining estimated click rates of the related searches respectively according to a click estimating model obtained from pre-training, wherein the determining includes considering an obtained user behavior feature as input of the click estimating model to obtain an estimated click rate of a related search output by the click estimating model;
sending the estimated click rates of the related searches from a first processing subunit to a second processing subunit to respectively determine guiding capability scores of the related searches according to the estimated click rates and guiding capability features of the related searches;

sorting the related searches based on the estimated click rates that are determined according to the click estimating model, wherein the sorting of the related searches based on the estimated click rates comprises:
sorting, by a sorting subunit of the computer, all the related searches in a descending order according to the guiding capability scores, wherein the respective determination of the guiding capability scores of the related searches according to the estimated click rates and the guiding capability features of the related searches comprises:
during a user's searching, performing, by the second processing subunit of the computer, following operation for each related search of the related searches:
respectively obtaining click times of the related searches displayed within a recent predetermined time length after the user clicks the related search, adding up the click times, dividing a sum by times of displaying the related searches within the recent predetermined time length after the related search is clicked, to obtain a guiding capability feature of the related search;
obtaining a click rate of the related search within the recent predetermined time length; and
calculating a product of the guiding capability feature of the related search and the click rate of the related search within the recent predetermined time length, and adding up the product with the estimated click rate of the related search after having gone through normalization processing to obtain a guiding capability score of the related search; and

after sorting all the related searches in the descending order according to the guiding capability scores and the estimated click rates to improve accuracy of the sorting of all the related searches, displaying, by the computer, top N related searches in a search result page, and recommending the displayed top N related searches to a user, the N is a positive integer.

6. The device according to claim 5 , wherein the determining estimated click rates of the related searches respectively according to a click estimating model obtained from pre-training comprises:
for each related search, obtaining a user behavior feature of the related search respectively.

7. The device according to claim 6 , wherein the user behavior feature comprises: co-occurrence feature, quality control feature, click feature and semantic feature.

8. The device according to claim 5 , wherein the respectively determining quality scores of related searches according to quality features of related searches comprises:
as for each related search, determining the quality score of the related search in a fitting manner respectively according to the quality control feature, guiding capability feature and estimated click rate of the related search.

9. A non-transitory computer storage medium, wherein the computer storage medium is coded with a computer program, and when the program is executed by one or more computers, the one or more computers perform following operation:
obtaining, by

Assignees

• Baidu online network technology beijing co ltd

Inventors

• Wang Wei
• Gui jia
• Cheng Jun
• Wang Kuipeng

Classifications

• G06F16/24578 Primary

  using ranking · CPC title

• G06F7/08

  Sorting, i.e. grouping record carriers in numerical or other ordered sequence according to the classification of at least some of the information they carry (by merging two or more sets of carriers in ordered sequence G06F7/16) · CPC title