A computer based agent may employ a variety of techniques including machine learning to analyze queries to a database, improve the queries to a database and make the improved queries available to new and old user through a user interface.

Described herein is a technique to facilitate conjunctive filtering for an embedding-based information retrieval system. Each item in a content collection is initially analyzed to identify its property values for some set of properties. For each item, its property values are encoded in the form of a vector, and concatenated with an item embedding for the item to derive an enhanced item embedding. When a query is received, a query embedding is generated. Any filtering parameters received with or as part of the query are encoded in the form of a vector, and the vector is concatenated with the query embedding to derive an enhanced query embedding. The enhanced query embedding is then used in a “k” approximate nearest neighbor search to identify items relevant to the query and having property values that satisfy the filtering parameters.

Techniques are disclosed relating to methods that include preprocessing, by a computer system, records of a database to create one or more token sets for a given record. The created token sets may correspond to ones of a plurality of search string functions, and may include token sets that include a plurality of possible substrings located within data strings of a corresponding database record. The methods may further include receiving a query for a search of the database. The query may include at least one of the plurality of search string functions. The method may also include performing the search by traversing, using at least a portion of the records, at least one token set corresponding to the included search string functions, as well as returning results for the search based on the query and the traversing.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving a query provided by a user and comprising one or more terms. Obtaining context data based on at least a portion of a first resource displayed to the user at a time that the query is received. Obtaining a revised query that is based on the query and the context data. Receiving a plurality of search results responsive to the revised query. Automatically, selecting a search result that represents a second resource from the plurality of search results, and providing the second resource for display to the user.

Methods and systems disclosed herein may optimize write operations in a transaction to reduce the number of operations to the point where each of the write operations may be performed in parallel. The writer optimizer may review a first write operation and a second write operation to determine whether the first write operation and the second write operation write to the same row in the same table. When the first write operation and the second write operation write to the same row in the same table, the first write operation and the second write operation may be combined to generate a functionally-equivalent third write operation. The third write operation may reduce the number of operations in the transaction by replacing both the first write operation and the second write operation.

Systems for dynamic data reorganization and query optimization are provided. In some aspects, a database query may be received and evaluated to determine whether it meets one or more criteria. If the database query meets the one or more criteria, the query may be executed and the output of the executed database may be transmitted to the user device. If the database query does not meet the one or more criteria, data associated with the database query may be dynamically modified and/or reorganized (e.g., using machine learning). Modified data may be generated based on the dynamically modified and/or reorganized data. The modified data may be stored in an optimized query cache. An updated query may be generated based on the modified data and stored in the optimized query cache. The updated query may be executed and an output of the executed updated query may be transmitted to the user device.

The lock time of a database upon the execution of a series of processes is decreased, and the load of the implementation/modification of the database server device upon changing the application destination of the Web system is reduced. A database server provides a prescribed simple API, and comprises a request processing unit, an SQL creation unit, and an SQL processing unit. In the database server device 120, when a plurality of divided requests are received from the Web server device as the request for requesting a series of processing, the SQL processing unit executes SQL integration processing of temporarily storing the generated SQL sentences until generation of the SQL sentences is ended for all of the plurality of divided requests, and integrating and editing the plurality of temporarily stored SQL sentences and creating the whole SQL sentence after generation of the SQL sentence is ended for all of the requests.

A computer-implemented method provides application-based query transformations. The method includes determining an application is initiated. The method includes identifying a set of execution units included in the application. The execution units are based on of a set of queries in the application and a set of actions in the application. The method also includes building a query dependency graph (QDG) comprising a plurality of nodes, wherein each node of the plurality of nodes is correlated to an execution unit, and each node is linked to at least one additional node, the link indicating a relative execution order and a common attribute each node and the additional node. The method includes merging, based on a performance architecture, two or more of the set of execution units into a section. The method includes processing the application according to the QDG.

A computer-implemented method masks shard operations in a distributed database. The method includes performing a shard operation on a first table with a first column and a second column, where the shard operation divides the first table into a first shard with the first column and a second shard with the second column. The method also includes receiving a query, where the query is configured to perform at least a first command. The method further includes determining the first command is directed to the first column. The method includes updating the query to access the first shard in place of the first table. The method also includes executing the query and returning a set of results for the query to a source of the query.

Events from one or more primary systems associated with one or more tenants are received. The received events are stored in a message queue. At least a portion of the events in the message queue are ingested for organization and storage in a data store. One or more progress identifiers associated with ingesting of the events in the message queue are tracked. An event query is received from an external system. The event query is rewritten into a first component query for the data store and a second component query for the message queue based at least in part on a progress identifier. A result of the first component query and a result of the second component query are combined to determine a result of the event query.