Identification and issuance of repeatable queries
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, that identify and issue search queries expected to be issued in the future. A set of search queries that have been issued by multiple user devices can be obtained. For each query instance, contextual data can be obtained. A first query and its contextual data can be input to a model that outputs the query's likelihood of being issued in the future. The model can be trained using contextual data for training queries and a corresponding labels for the training queries. The learning model outputs the first query's likelihood of being issued in future, and this query is stored as a repeatable query if the likelihood satisfying a repeatability threshold. Subsequently, a stored repeatable query is issued upon a selection of a user selectable interface component and the search engine provides search results for the query.
1. A computer-implemented method, the method comprising:
obtaining, from a search engine query log and by a computing system comprising one or more processors, a plurality of training queries that have been issued by different user devices;
for each of the plurality of training queries:
determining and storing in the search engine query log, with a context analyzer and based on a context in which a respective training query was issued and user interactions with search results pages in response to the respective training query, a respective contextual dataset, wherein the user interactions are determined by accessing a click log;
obtaining, by the computing system, training dataset, wherein the training dataset comprises a plurality of contextual datasets, a plurality of labels, and the plurality of training queries, wherein each label indicates whether the respective training query was observed to be repeated or otherwise to be repeatable, wherein each of the plurality of contextual datasets comprise contextual data determined for the respective training query of the plurality of training queries, and wherein each of the plurality of labels indicates whether the respective training query has been issued a threshold number of times, wherein each of the plurality of contextual datasets for the plurality of training queries comprises: a number of times that a particular query has been issued by a particular user device and a number of unique user devices that issued the particular query a threshold number of times;
processing, by the computing system, a contextual dataset of the plurality of contextual datasets with a learning model to determine a likelihood that a first training query associated with the contextual dataset will be issued in the future;
generating and storing in storage, by the computing system, a repeatable query determination as an output for the first training query based on the likelihood that the first query will be issued in the future and a repeatability threshold; and
training, by the computing system, the learning model based on the repeatable query determination for the first training query and an output of a label of the plurality of labels for the first training query.
2. The method of claim 1 , further comprising:
obtaining a set of search queries that have been issued by a plurality of user devices;
determining a set of respective contextual datasets for the set of search queries;
determining a plurality of repeatable queries based on processing the set of respective contextual datasets with the learning model to determine a plurality of respective likelihoods; and
storing the plurality of repeatable queries with other repeatable search queries that have been previously identified as repeatable queries.
3. The method of claim 2 , wherein storing the plurality of repeatable queries with other repeatable search queries that have been previously identified as repeatable queries comprises storing, by the search engine, the plurality of repeatable queries with search results responsive to the plurality of repeatable queries.
4. The method of claim 2 , wherein each of the set of respective contextual datasets for the set of search queries, comprises:
a language in which a particular query is written;
a geographic location from which the particular query is issued; and
a geographic location of interest to the user device that issued the particular query.
5. The method of claim 2 , wherein each of the set of respective contextual datasets for the set of search queries, comprises at least one of:
an embedding of a particular query that represents a semantic relationship between the particular query and other queries; or
a determination as to whether the particular query is directed to a particular web location or website.
6. The method of claim 2 , further comprising:
providing, on a user device, a user selectable interface component that, upon being selected by a user device and without receiving a user input of a component of a query, results in issuance of a query from among the repeatable queries;
receiving, from the user device, a first selection of the user selectable interface component that requests issuance of a particular query from among the repeatable queries; and
providing, by a search engine and in response to receiving the first selection from the user device, a first search results page including search results for the particular query.
7. The method of claim 6 , further comprising:
receiving, from the user device, a second selection of the user selectable interface component that requests the particular query to be issued; and
providing, by the search engine and in response to receiving the second selection from the user device, a second search results page including search results for the particular query, wherein the search results page is different from the first search results page.
8. The method of claim 7 , wherein the second search results page is different from the first search results page when:
search results of the second results page are ordered differently from the search results of the first search results page;
search results of the second results page are different from the search results of the first search results page; or
the second results page includes dynamic content that is not included on the first results page.
9. The method of claim 8 , wherein the dynamic content comprises content generated more recently than content from the first search results page.
10. A computing system, the system comprising:
one or more processors;
one or more non-transitory computer readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:
obtaining, from a search engine query log, a plurality of training queries that have been issued by different user devices;
for each of the plurality of training queries:
determining and storing in the search engine query log, with a context analyzer and based on a context in which a respective training query was issued and user interactions with search results pages in response to the respective training query, a respective contextual dataset, wherein the user interactions are determined by accessing a click log;
obtaining training dataset, wherein the training dataset comprises a plurality of contextual datasets, a plurality of labels, and the plurality of training queries, wherein each label indicates whether the respective training query was observed to be repeated or otherwise to be repeatable, wherein each of the plurality of contextual datasets comprise contextual data determined for the respective training query of the plurality of training queries, and wherein each of the plurality of labels indicates whether the respective training query has been issued a threshold number of times, wherein each of the plurality of contextual datasets for the plurality of training queries comprises: a number of times that a particular query has been issued by a particular user device and a number of unique user devices that issued the particular query a threshold number of times;
processing a contextual dataset of the plurality of contextual datasets with a learning model to determine a likelihood that a first training query associated with the contextual dataset will be issued in the future;
generating and storing in storage a repeatable query determination as an output for the first training query based on the likelihood that the first query will be issued in the future and a repeatability threshold;
and training the learning model based on the repeatable query determination for the first training query and an output of a label of the plurality of labels for the first training query.
11. The system of claim 10 , wherein the operations further comprise:
obtaining a search query;
determining respective contextual data for the search query; and
processing the respective contextual data with the learning model to determine a respective likelihood that the search query will be issued in the future.
12. The system of claim 11 , wherein the contextual data represents a context in which the search query was issued and user interactions with search results pages provided in response to the search query.
13. The system of claim 11 , wherein the operations further comprise:
identifying the search query as a repeatable query based on the respective likelihood that the search query will be issued in the future satisfying a repeatability threshold;
storing the search query with other repeatable search queries that have been previously identified as repeatable queries; and
providing, on a user device, a user selectable interface component that, upon being selected by a user device and without receiving a user input of a component of a query, results in issuance of a query from among the repeatable queries.
14. The system of claim 13 , wherein the operations further comprise:
receiving, from the user device, a first selection of the user selectable interface component that requests issuance of a particular query from among the repeatable queries; and
providing, by a search engine and in response to receiving the first selection from the user device, a first search results page including search results for the particular query.
15. One or more non-transitory computer readable media that collectively store instructions that, when executed by one or more processors, cause a computing system to perform operations, the operations comprising:
obtaining, from a search engine query log, a plurality of training queries that have been issued by different user devices; for each of the plurality of training queries:
determining and storing in the search engine query log, with a context analyzer and based on a context in which a respective training query was issued and user interactions with search results pages in response to the respective training query, a respective contextual dataset, wherein the user interactions are determined by accessing a click log;
obtaining training dataset, wherein the training dataset comprises a plurality of contextual datasets, a plurality of labels, and the plurality of training queries, wherein each label indicates whether the respective training query was observed to be repeated or otherwise to be repeatable, wherein each of the plurality of contextual datasets comprise contextual data determined for the respective training query of the plurality of training queries, and wherein each of the plurality of labels indicates whether the respective training query has been issued a threshold number of times, wherein each of the plurality of contextual datasets for the plurality of training queries comprises: a number of times that a particular query has been issued by a particular user device and a number of unique user devices that issued the particular query a threshold number of times;
processing a contextual dataset of the plurality of contextual datasets with a learning model to determine a likelihood that a first training query associated with the contextual dataset will be issued in the future;
generating and storing in storage a repeatable query determination as an output for the first training query based on the likelihood that the first query will be issued in the future and a repeatability threshold;
and training the learning model based on the repeatable query determination for the first training query and an output of a label of the plurality of labels for the first training query.
16. The one or more non-transitory computer readable media of claim 15 , wherein the contextual dataset comprises: a selection by a user device of one or more search results provided on a search results page for a respective query.
17. The one or more non-transitory computer readable media of claim 15 , wherein the contextual dataset comprises: a time of viewing by a user device of one or more search results provided on a search results page for a respective query.
18. The one or more non-transitory computer readable media of claim 15 , wherein the contextual dataset comprises: a selection of navigational interface elements on a search results page provided for a respective query.
19. The one or more non-transitory computer readable media of claim 15 , wherein the learning model comprises a machine learning model and a rules engine.