System and method for detecting a search intent

• US 10,191,991 B2
• Filed: 12/12/2016
• Issued: 01/29/2019
• Est. Priority Date: 06/11/2010
• Status: Expired due to Fees

• Current Assignee: DoAT Media Ltd.
• Original Assignee: DoAT Media Ltd.
• Inventors: Kasterstein, Rami, Ben-David, Amihay, Simhon, Joey Joseph
• Primary Examiner(s): ASPINWALL, EVAN S
• Application Number:
  US15/375,844
• Publication Number:
  US 20170091338A1
• Time in Patent Office: 778 Days
• Field of Search: 707706
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/34 Browsing; Visualisation therefor
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/953 Querying, e.g. by the use of web search engines
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9538 Presentation of query results

Abstract

A method and system for detecting a search intent of a user. The method includes tokenizing the search query into at least one tokenized query; sending, to a plurality of engines, the at least one tokenized query, wherein each engine is configured with at least one entity, wherein each entity represents a topic of interest; receiving, from each engine, the at least one entity of the engine and at least one certainty score, wherein each certainty score indicates a probability that a tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine; and determining, based on the received entities and certainty scores, the search intent of the user.

Claims

1. A method for determining a search intent of a user submitting a search query, comprising:

tokenizing the search query into at least one tokenized query, wherein the tokenizing further comprises creating a plurality of tokens based on the search query and arranging the created tokens based on a graph of connections between entities;

transmitting, to a plurality of engines, the at least one tokenized query, wherein each engine is configured with at least one entity, wherein each entity represents a topic of interest;

receiving, from each of at least two engines of the plurality of engines, the at least one entity of the engine and at least one certainty score, wherein each certainty score indicates a probability that a tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine;

determining, based on the received entities and their associated certainty scores, the search intent of the user; and

supplying the determined search intent of the user to a search server for use in developing search results for the user submitted search query.

2. The method of claim 1, wherein each received certainty score is above a predetermined threshold.

3. The method of claim 1, wherein determining the search intent of the user further comprises:

statistically analyzing the received entities and certainty scores, wherein the statistical analysis includes determining a co-occurrence of the at least one tokenized query within the at least one entity of each engine, wherein the search intent of the user is determined further based on the determined co-occurrences.

4. The method of claim 1, wherein determining the search intent of the user further comprises:

semantically analyzing the at least one tokenized query, wherein the semantic analysis includes determining a combination of the at least one tokenized query that describes a coherent query.

5. The method of claim 1, wherein the search query is any of:

a definitive query, a fuzzy query, and a noisy query.

6. The method of claim 5, wherein tokenizing the search query into at least one tokenized query further comprises:

filtering, based on a list of unnecessary tokens, at least one unnecessary token from the search query, when the search query is a fuzzy query or a noisy query.

7. The method of claim 1, wherein the tokenizing is based on a graph of connections between entities of the plurality of engines.

8. The method of claim 1, further comprising:

receiving at least one environmental variable and at least one personal variable;

analyzing each of the at least one personal variable and the at least one environmental variable to determine a context;

correlating each of the received at least one personal variable and at least one environmental variable based on their respective contexts; and

deriving, based on the correlation, at least a category of interest, wherein the search intent is determined based further on the category of interest.

9. The method of claim 1, wherein each engine is configured to compute the probability that each tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine.

10. A non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to perform a process, the process comprising:

tokenizing the search query into at least one tokenized query, wherein the tokenizing further comprises creating a plurality of tokens based on the search query and arranging the created tokens based on a graph of connections between entities;

transmitting, to a plurality of engines, the at least one tokenized query, wherein each engine is configured with at least one entity, wherein each entity represents a topic of interest;

receiving, from each of at least two engines of the plurality of engines, the at least one entity of the engine and at least one certainty score, wherein each certainty score indicates a probability that a tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine;

determining, based on the received entities and their associated certainty scores, the search intent of the user; and

supplying the determined search intent of the user to a search server for use in developing search results for the user submitted search query.

11. An intent detector for determining a search intent of a user submitting a search query, comprising:

a network interface, wherein the network interface is configured to at least receive the search query;

a processing circuitry; and

a memory connected to the processing circuitry and configured to contain a plurality of instructions that, when executed by the processing circuitry, configure the intent detector to;

tokenize the search query into at least one tokenized query, wherein the system is further configured to;

create a plurality of tokens based on the search query and arrange the created tokens based on a graph of connections between entities;

transmit, to a plurality of engines, the at least one tokenized query, wherein each engine is configured with at least one entity, wherein each entity represents a topic of interest;

receive, from each of at least two engines of the plurality of engines, the at least one entity of the engine and at least one certainty score, wherein each certainty score indicates a probability that a tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine;

determine, based on the received entities and their associated certainty scores, the search intent of the user; and

supply the determined search intent of the user to a search server for use in developing search results for the user submitted search query.

12. The intent detector of claim 11, wherein each received certainty score is above a predetermined threshold.

13. The intent detector of claim 11, wherein the intent detector is further configured to:

statistically analyze the received entities and certainty scores, wherein the statistical analysis includes determining a co-occurrence of the at least one tokenized query within the at least one entity of each engine, wherein the search intent of the user is determined further based on the determined co-occurrences.

14. The intent detector of claim 11, wherein the intent detector is further configured to:

semantically analyze the at least one tokenized query, wherein the semantic analysis includes determining a combination of the at least one tokenized query that describes a coherent query.

15. The intent detector of claim 11, wherein the search query is any of:

a definitive query, a fuzzy query, and a noisy query.

16. The intent detector of claim 15, wherein the intent detector is further configured to:

filter, based on a list of unnecessary tokens, at least one unnecessary token from the search query, when the search query is a fuzzy query or a noisy query.

17. The intent detector of claim 11, wherein the tokenizing is based on a graph of connections between entities of the plurality of engines.

18. The intent detector of claim 11, wherein the intent detector is further configured to:

receive at least one environmental variable and at least one personal variable;

analyze each of the at least one personal variable and the at least one environmental variable to determine a context;

correlate each of the received at least one personal variable and at least one environmental variable based on their respective contexts; and

derive, based on the correlation, at least a category of interest, wherein the search intent is determined based further on the category of interest.

19. The intent detector of claim 11, wherein each engine is configured to compute the probability that each tokenized query of the at least one tokenized query is mapped to the at least one entity of the engine.