Extended search method and apparatus

Patent metadata

Field	Value
Publication number	US-2018039675-A1
Application number	US-201615393734-A
Country	US
Kind code	A1
Filing date	Dec 29, 2016
Priority date	Aug 4, 2016
Publication date	Feb 8, 2018
Grant date	—

Abstract

An extended search method and apparatus is provided. An interest term model of each user is established. An extended term model of each fourth search sequence is established. A corresponding extended search term is determined based on a current search sequence of a current user and based on an interest term model of the current user as well as an extended term model of the current search sequence; and a corresponding search result is provided for the current user based on the current search sequence and the extended search term. The present invention can implement simpler and more efficient personalized searches, and is advantageous in terms of meeting long-tail requirements of users (that is, adding personalized result recalling), reducing calculation and storage costs, universality, practicability, and the like.

First claim

What is claimed is:

1 . An extended search method, comprising the following steps:
establishing an interest term model of each user;
wherein an interest search term set of the each user is generated based on a first search sequence submitted by the each user in a predetermined time window and a second search sequence associated with a browsing search result corresponding to the first search sequence, to establish the interest term model of the each user;
establishing an extended term model of each fourth search sequence, wherein
based on all third search sequences in the predetermined time window and fourth search sequences having a length less than a predetermined threshold among the third search sequences, fifth search sequences comprising the fourth search sequences are acquired from the third search sequences; and
for the each fourth search sequence, an extended term set of the each fourth search sequence is generated based on a differential part between the each fourth search sequence and the fifth search sequence corresponding to the each fourth search sequence, to establish the extended term model of the each fourth search sequence;
wherein the method further comprises:
determining a corresponding extended search term based on a current search sequence of a current user and based on an interest term model of the current user and an extended term model of the current search sequence; and
providing a corresponding search result for the current user based on the current search sequence and the extended search term.

2 . The method according to claim 1 , wherein the interest term model of the each user is established based on interest search terms in the interest search term set of the corresponding user and weights of the interest search terms; and the extended term model of the each fourth search sequence is established based on extended terms in the extended term set of the corresponding fourth search sequence and weights of the extended terms;
wherein the step of determining the extended search term further comprises:
determining the extended search term based on an identical term in an interest search term set of the current user and an extended term set of the current search sequence and a ranking weight of the identical term.

3 . The method according to claim 1 , wherein the interest term model of the each user is established based on classes of interest search terms in the interest search term set of the corresponding user; and the extended term model of the each fourth search sequence is established based on classes of extended terms in the extended term set of the corresponding fourth search sequence;
wherein the step of determining the extended search term further comprises:
determining a candidate extended class based on an identical class corresponding to an interest search term set of the current user and an extended term set of the current search sequence and a ranking weight of the identical class; and
selecting the extended search term from terms in the candidate extended class.

4 . The method according to claim 3 , wherein the step of establishing the interest term model further comprises:
classifying the interest search terms in the interest search term set of the each user, to obtain multiple classes, a weight of each class, and interest search terms included in the each class, thereby establishing the interest term model of the corresponding user; wherein the step of establishing the extended term model further comprises:
classifying the extended terms in the extended term set of the each fourth search sequence, to obtain multiple classes, a weight of the each class, and extended terms comprised in the each class, thereby establishing the extended term model of the corresponding fourth search sequence.

5 . The method according to claim 1 , wherein the interest term model of the each user is established based on a user-interest search term matrix; and the extended term model of the each fourth search sequence is established based on a fourth search sequence-extended term matrix;
wherein the step of determining the extended search term further comprises:
acquiring an interest search term corresponding to the current user based on the user-interest search term matrix, acquiring an extended term corresponding to the current search sequence based on the fourth search sequence-extended term matrix, and using the interest search term and the extended term as a candidate extended search term set; and
selecting the extended search term from the candidate extended search term set based on a degree of interest of each term in the candidate extended search term set with respect to the current user and a degree of correlation between the each term and the current search sequence.

6 . The method according to claim 5 , wherein the step of establishing the interest term model further comprises:
establishing the user-interest search term matrix based on the each user and the interest search term set of the user; and determining, based on the user-interest search term matrix, a degree of interest of the each user on each interest search term in the user-interest search term matrix, thereby establishing the interest term model of the corresponding user; wherein the step of establishing the extended term model further comprises:
establishing the fourth search sequence-extended term matrix based on the each fourth search sequence and the extended term set of the each fourth search sequence; and
determining, based on the fourth search sequence-extended term matrix, a degree of correlation between the each fourth search sequence and the each extended term in the fourth search sequence-extended term matrix, thereby establishing the extended term model of the corresponding fourth search sequence.

7 . The method according to claim 1 , wherein the interest term model of the each user is established based on a user-interest search term matrix;
wherein the step of determining the extended search term further comprises:
determining, based on the user-interest search term matrix, degrees of interest of the current user on extended terms in an extended term set of the current search sequence, to select the extended search term from the extended term set.

8 . The method according to claim 7 , wherein the step of selecting the extended search term further comprises:
selecting the extended search term from the extended term set of the current search sequence based on the degrees of interest of the current user on the extended terms in the extended term set in combination with weights of the extended terms in the extended term set.

9 . The method according to claim 7 , wherein the step of selecting the extended search term further comprises:
determining a candidate extended term class based on weights of extended term classes in the extended term set of the current search sequence; and selecting the extended search term from the candidate extended term class based on degrees of interest of the current user on extended terms in the candidate extended term class.

10 . The method according to claim 1 , wherein the extended term model of the each fourth search sequence is established based on a fourth search sequence-extended term matrix;
wherein the step of determining the extended search term further comprises:
determining degrees of correlation between the current search sequence and interest search terms in an interest search term set of the current user based on the fourth search sequence-extended term matrix, to select the extended search term from the interest search term set.

11 . The method according to claim 10 , wherein the step of

Assignees

• Baidu online network technology beijing co ltd

Inventors

• Gu Xiaoguang
• Fang Gaolin

Classifications

• G06F16/2457 Primary

  with adaptation to user needs · CPC title

• G06F16/583

  using metadata automatically derived from the content · CPC title

• G06F16/2455

  Query execution · CPC title

• G06F17/30522 Primary

  Physics · mapped topic

• G06F17/30477

  Physics · mapped topic