SYSTEM FOR AUTOMATICALLY CATEGORIZING QUERIES

• US 20100094854A1
• Filed: 02/09/2009
• Published: 04/15/2010
• Est. Priority Date: 10/14/2008
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Rouhani-Kalleh, Omid
• Granted Patent: US 8,041,733 B2
• Time in Patent Office: Days
• US Class Current: 707/706
• CPC Class Codes: G06F 16/3325 Reformulation based on results of preceding query

Abstract

A system and method is provided for recognizing a pattern in queries and using the recognized pattern to categorize queries under existing query categories. An entity text that represents a real-world object is detected in a query. The entity text is categorized into an entity category. The query is then rewritten as an annotated query by replacing the entity text with a placeholder for any entity matching the entity category. The annotated query is compared to other queries that are associated with query categories to find a query category that is frequently associated with the annotated query. The query category frequently associated with the annotated query is used to generate a rule. The rule indicates that any query matching the annotated query is likely to fit into the query category. The rule can be used to assign categories to queries or correct incorrectly categorized queries.

Claims

1. A computer-implemented method comprising:

retrieving, from a volatile or non-volatile computer-readable storage medium, a query associated with a query category;

detecting an entity text in the query;

mapping the entity text to an entity category;

determining a keyword text from a remainder of the query;

determining that another query comprises said keyword text and another entity text in said entity category;

based at least in part on said determining that another query comprises said keyword text and another entity text in said entity category, storing, on a volatile or non-volatile computer-readable storage medium, information that indicates that the other query is associated with said query category.

2. A computer-implemented method comprising:

retrieving, from a volatile or non-volatile computer-readable storage medium, a set of queries and a set of query categories, each query of the set of queries associated with a query category of the set of query categories;

for each query;

detecting an entity text in the query;

mapping the entity text to an entity category;

determining a keyword text from a remainder of the query;

storing, on a volatile or non-volatile computer readable storage medium, information that indicates that an annotated query comprising the keyword text and the entity category is associated with the query category.

3. The computer-implemented method of claim 2, wherein said information is a first information, further comprising:

determining a particular annotated query of a particular keyword text and a particular entity category that is frequently associated with a particular query category;

storing, on the volatile or non-volatile computer-readable storage medium, a second information that indicates that another query is associated with the particular query category when the other query comprises the particular keyword text and another entity text in the particular entity category.

4. The computer-implemented method of claim 2, further comprising:

for each annotated query, ranking the annotated query based at least in part on a frequency by which the pair occurs in the set of queries; and

determining the annotated queries that are above a threshold rank.

5. The computer-implemented method of claim 2, further comprising:

for each annotated query, ranking the annotated query based at least in part on a frequency by which the annotated query occurs relative to a total number of times that the entity category occurred in the set of queries; and

determining the annotated queries that are above a threshold rank.

6. The computer-implemented method of claim 2, further comprising:

for each annotated query, ranking the annotated query based at least in part on a frequency by which the annotated query occurs relative to a total number of times that the keyword text occurred in the set of queries; and

determining the annotated queries that are above a threshold rank.

7. The computer-implemented method of claim 2, further comprising:

for each annotated query, ranking the annotated query based at least in part on a frequency by which the annotated query occurs in unique queries of the set of queries; and

determining the annotated queries that are above a threshold rank.

8. The computer-implemented method of claim 2, further comprising:

for each annotated query, determining an accuracy value for the annotated query based at least in part on the frequency by which the annotated query refers to the query category relative to the total number of times the annotated query occurs in the set of queries;

determining the annotated queries that are above a threshold accuracy value.

9. The computer-implemented method of claim 2, further comprising storing the annotated query by replacing the entity text in the query with a placeholder for any entity text in the entity category.

10. The computer-implemented method of claim 2, further comprising:

detecting a word in the query that belongs to a list of generic words; and

replacing the word with a placeholder for any keyword text.

11. The computer-implemented method of claim 2, wherein said entity text is a first entity text, and said entity category is a first entity category, further comprising:

for each query;

detecting a second entity text in the query;

mapping the second entity text to a second entity category;

wherein the annotated query further comprises the second entity category; and

wherein said storing the information that indicates that an annotated query is associated with said query category occurs when the other query comprises;

said keyword text,

a third entity text in said first entity category, and

a fourth entity text in said second entity category.

12. The computer-implemented method of claim 11, further comprising storing the annotated query by:

replacing the first entity text in the query with a first placeholder for any entity text in the first entity category; and

replacing the second entity text in the query with a second placeholder for any entity text in the second entity category.

13. A computer-implemented method comprising:

receiving a query;

detecting an entity text in the query;

mapping the entity text to an entity category;

determining a keyword text from a remainder of the query;

determining that the query matches an annotated query comprising said entity category and said keyword text, the annotated query associated with a query category;

based on said determining that the query matches the annotated query, storing, on a volatile or non-volatile computer-readable storage medium, information that indicates that the query is associated with said query category.

14. A computer-implemented method comprising:

receiving a query;

detecting a first entity text in the query;

mapping the first entity text to a first entity category;

determining a first keyword text for the query;

detecting a second entity text in the query;

mapping the second entity text to a second entity category;

determining a second keyword text for the query;

determining that the query matches a first annotated query comprising the first entity category and the first keyword text, the first annotated query associated with a first query category;

determining that the query matches a second annotated query comprising the second entity category and the second keyword text, the second annotated query associated with a second query category;

determining whether the first query category matches the second query category;

in response to said determining whether the first query category matches the second query category, modifying, on a volatile or non-volatile computer-readable storage medium, an accuracy value by which the query is associated with the first query category.

15. The computer-implemented method of claim 14, further comprising:

storing a list of bits for each query category, the list of bits comprising at least one bit for each annotated query;

in response to determining that the query matches a first annotated query, storing 1 in a first bit associated with the first annotated query in the list of bits; and

in response to determining that the query matches a second annotated query, storing 1 in a second bit associated with the second annotated query in the list of bits.