Contextual query revision
View Patent ↗Apparatus, systems and methods for contextual query revision are disclosed. A current search query is received during a search session. The current search query includes one or more current search tokens. Potentially inaccurate search tokens are identified from the one or more current search tokens. A possible replacement token is identified based upon the potentially inaccurate search token. A group of related tokens is identified from query logs, and a modified search query is generated if the replacement token is not included in the related tokens.
1. A method, comprising:
receiving a first set of search queries during a first search session;
identifying, from the first set of search queries, a first set of search tokens;
identifying, from a query log storing search queries for a plurality of different search sessions, a second search session during which at least a threshold portion of the first set of search tokens were included in a second set of search queries received during the second search session;
selecting, by one or more computers and as related tokens for the first search session, a second set of search tokens from the second set of search queries;
receiving a current search query during the first search session;
determining, by one or more computers and based on the query log, that the current search query has historically been revised by users to replace a particular search token from the first set of search tokens with one of the related tokens; and
creating, by one or more computers, a suggested search query in which the particular search token is replaced by the one of the related tokens.
2. The method of claim 1 , comprising providing the suggested search query to a user device in response to receipt of the current search query.
3. The method of claim 1 , wherein selecting related tokens for the first session comprises selecting, as the related search tokens, search tokens that were included in at least a threshold portion of search queries from the second set of search queries.
4. The method of claim 1 , wherein selecting related tokens for the first search session comprises:
clustering search tokens from the second search session into token clusters; and
selecting two or more of the related tokens from a same token cluster.
5. The method of claim 4 , wherein clustering the search tokens from the second search session comprises:
identifying, for each search token in the second set of search queries, search queries from the second set of search queries that include the search token; and
creating the token clusters to include only search tokens for which a threshold number of the identified search queries also included another search token in the cluster.
6. The method of claim 1 , comprising:
determining a context for the current search session based on subject weights for the current search query and subject weights for search queries in the first set of search queries; and
selecting, as a replacement token for the particular search token, the related token having a highest subject weight for the context.
7. The method of claim 1 , comprising obtaining search results based on the suggested search query.
8. The method of claim 1 , comprising selecting a replacement token for the particular search token from the set of related tokens.
9. The method of claim 8 , wherein selecting a replacement token comprises selecting, as the replacement token, a related token from the set of related tokens based on a distance vector between the particular token and the related token.
10. The method of claim 9 , wherein selecting a replacement token comprises selecting, as the replacement token, a related token from the set of related tokens based on a distance vector between the particular token and the related token and a subject weight of the related token.
11. A system comprising:
one or more data stores storing a query log that includes search queries for a plurality of different search sessions; and
data processing apparatus including one or more processors that execute instructions that cause the one or more processors to interact with the data store and to perform actions including:
receiving a first set of search queries during a first search session;
identifying, from the first set of search queries, a first set of search tokens;
identifying, from the query log, a second search session during which at least a threshold portion of the first set of search tokens were included in a second set of search queries received during the second search session;
selecting, as related tokens for the first search session, a second set of search tokens from the second set of search queries;
receiving a current search query during the first search session;
determining, based on the query log, that the current search query has historically been revised by users to replace a particular search token from the first set of search tokens with one of the related tokens; and
creating a suggested search query in which the particular search token is replaced by the one of the related tokens.
12. The system of claim 11 , wherein the instructions cause the one or more processors to perform actions comprising providing the suggested search query to a user device in response to receipt of the current search query.
13. The system of claim 11 , wherein selecting related tokens for the first session comprises selecting, as the related search tokens, search tokens that were included in at least a threshold portion of search queries from the second set of search queries.
14. The system of claim 11 , wherein selecting related tokens for the first search session comprises:
clustering search tokens from the second search session into token clusters; and
selecting two or more of the related tokens from a same token cluster.
15. The system of claim 14 , wherein clustering the search tokens from the second search session comprises:
identifying, for each search token in the second set of search queries, search queries from the second set of search queries that include the search token; and
creating the token clusters to include only search tokens for which a threshold number of the identified search queries also included another search token in the cluster.
16. The system of claim 11 , wherein the instructions cause the one or more processors to perform actions comprising:
determining a context for the current search session based on subject weights for the current search query and subject weights for search queries in the first set of search queries; and
selecting, as a replacement token for the particular search token, the related token having a highest subject weight for the context.
17. The system of claim 11 , wherein the instructions cause the one or more processors to perform actions comprising obtaining search results based on the suggested search query.
18. The system of claim 11 , wherein the instructions cause the one or more processors to perform actions comprising selecting a replacement token for the particular search token from the set of related tokens.
19. The system of claim 18 , wherein selecting a replacement token comprises selecting, as the replacement token, a related token from the set of related tokens based on a distance vector between the particular token and the related token.
20. The system of claim 19 , wherein selecting a replacement token comprises selecting, as the replacement token, a related token from the set of related tokens based on a distance vector between the particular token and the related token and a subject weight of the related token.