Granted Patent B2

US 12,373,504 · App. 17/398,134 · Granted Jul 29, 2025

Inventor: Fuchun Jiang (Beijing, CN)

Assignee: Baidu Online Network Technology (Beijing) Co., Ltd.

G06F16/9535G06F16/951G06F16/9532

Abstract

Claims (54)

1. A method for recommending a search term, comprising:

acquiring an input target search term;

determining a neighbor node of a current node and a relationship between the current node and the neighbor node, by taking the target search term as the current node;

performing semantic representation processing on first text information of the current node and second text information of the neighbor node respectively, to obtain a first semantic representation of the current node and a second semantic representation of the neighbor node;

generating a cluster of the current node and the neighbor node based on the first semantic representation, the second semantic representation and the relationship between the current node and the neighbor node, and determining a semantic aggregation representation of the target search term by aggregating a semantic representation of the cluster; and

recalling a candidate search term to be recommended from a set of candidate search terms, by taking the semantic aggregation representation of the target search term as an index,

wherein determining the neighbor node of the current node by taking the target search term as the current node comprises acquiring display webpage information during searching, and determining a webpage title in the display webpage information as the neighbor node.

2. The method of claim 1 , further comprising:

screening the candidate search terms, according to statistical characteristics of search frequency of the target search term, result page display quantity and clicks; and

establishing the set of candidate search terms according to the screened candidate search terms.

3. The method of claim 2 , wherein the set of candidate search terms is present in a form of an approximate nearest neighbor (ANN) index library, and the establishing the set of candidate search terms according to the screened candidate search terms, comprises:

predicting semantic aggregation representations of all the candidate search terms, for all the screened candidate search terms; and

establishing the ANN index library, based on the semantic aggregation representations of all the candidate search terms.

4. The method of claim 3 , wherein the recalling the candidate search term to be recommended from the set of candidate search terms, by taking the semantic aggregation representation of the target search term as the index, comprises:

determining similarities between the semantic aggregation representations of all the candidate search terms and a semantic aggregation representation in the ANN index library, by taking the semantic aggregation representation of the target search term as the index;

determining top N semantic aggregation representations in similarity ranking, as target semantic aggregation representations, wherein N is a positive integer; and

recalling candidate search terms corresponding to the target semantic aggregation representations as the candidate search terms to be recommended.

5. An electronic device, comprising:

at least one processor; and

a memory communicatively connected with the at least one processor, wherein

the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform operations of:

acquiring an input target search term;

determining a neighbor node of a current node and a relationship between the current node and the neighbor node, by taking the target search term as the current node;

performing semantic representation processing on first text information of the current node and second text information of the neighbor node respectively, to obtain a first semantic representation of the current node and a second semantic representation of the neighbor node;

generating a cluster of the current node and the neighbor node based on the first semantic representation, the second semantic representation and the relationship between the current node and the neighbor node, and determining a semantic aggregation representation of the target search term by aggregating a semantic representation of the cluster; and

recalling a candidate search term to be recommended from a set of candidate search terms, by taking the semantic aggregation representation of the target search term as an index,

wherein determining the neighbor node of the current node by taking the target search term as the current node comprises acquiring display webpage information during searching, and determining a webpage title in the display webpage information as the neighbor node.

6. The electronic device of claim 5 , wherein the instructions are executable by the at least one processor to enable the at least one processor to further perform operations of:

screening the candidate search terms, according to statistical characteristics of search frequency of the target search term, result page display quantity and clicks; and

establishing the set of candidate search terms according to the screened candidate search terms.

7. The electronic device of claim 6 , wherein the set of candidate search terms is present in a form of an approximate nearest neighbor (ANN) index library, and the establishing the set of candidate search terms according to the screened candidate search terms, comprises:

predicting semantic aggregation representations of all the candidate search terms, for all the screened candidate search terms; and

establishing the ANN index library, based on the semantic aggregation representations of all the candidate search terms.

8. The electronic device of claim 7 , wherein the recalling the candidate search term to be recommended from the set of candidate search terms, by taking the semantic aggregation representation of the target search term as the index, comprises:

determining similarities between the semantic aggregation representations of all the candidate search terms and a semantic aggregation representation in the ANN index library, by taking the semantic aggregation representation of the target search term as the index;

determining top N semantic aggregation representations in similarity ranking, as target semantic aggregation representations, wherein N is a positive integer; and

recalling candidate search terms corresponding to the target semantic aggregation representations as the candidate search terms to be recommended.

9. A non-transitory computer-readable storage medium storing computer instructions for enabling a computer to perform operations of:

acquiring an input target search term;

determining a neighbor node of a current node and a relationship between the current node and the neighbor node, by taking the target search term as the current node;

performing semantic representation processing on first text information of the current node and second text information of the neighbor node respectively, to obtain a first semantic representation of the current node and a second semantic representation of the neighbor node;

generating a cluster of the current node and the neighbor node based on the first semantic representation, the second semantic representation and the relationship between the current node and the neighbor node, and determining a semantic aggregation representation of the target search term by aggregating a semantic representation of the cluster; and

recalling a candidate search term to be recommended from a set of candidate search terms, by taking the semantic aggregation representation of the target search term as an index,

wherein determining the neighbor node of the current node by taking the target search term as the current node comprises acquiring display webpage information during searching, and determining a webpage title in the display webpage information as the neighbor node.

10. The non-transitory computer-readable storage medium of claim 9 , wherein the computer instructions are executable by the computer to enable the computer to further perform operations of:

screening the candidate search terms, according to statistical characteristics of search frequency of the target search term, result page display quantity and clicks; and

establishing the set of candidate search terms according to the screened candidate search terms.

11. The non-transitory computer-readable storage medium of claim 10 , wherein the set of candidate search terms is present in a form of an approximate nearest neighbor (ANN) index library, and the establishing the set of candidate search terms according to the screened candidate search terms, comprises:

predicting semantic aggregation representations of all the candidate search terms, for all the screened candidate search terms; and

establishing the ANN index library, based on the semantic aggregation representations of all the candidate search terms.

12. The non-transitory computer-readable storage medium of claim 11 , wherein the recalling the candidate search term to be recommended from the set of candidate search terms, by taking the semantic aggregation representation of the target search term as the index, comprises:

determining similarities between the semantic aggregation representations of all the candidate search terms and a semantic aggregation representation in the ANN index library, by taking the semantic aggregation representation of the target search term as the index;

determining top N semantic aggregation representations in similarity ranking, as target semantic aggregation representations, wherein N is a positive integer; and

recalling candidate search terms corresponding to the target semantic aggregation representations as the candidate search terms to be recommended.

Assignments (2)

Priority Claims (1)

Continuity (1)

References Cited (15)