Salesforce Com, Inc.

Emc Corporation

Salesforce, Inc.

Francisco Borges

Marc Brette

Jayesh Govindarajan

Edward C Bueche

Venkatesan Chandrasekaran

Guillaume Kempf

Shanshan Quan

Petr Pleshachkov

Naren M Chittar

Anuprit Kale

Nicholas Beng Tek Geh

Edgar Gerardo Velasco

Ammar Haris

Guillaume Jean Mathieu Kempf

Arvind Srikantan

Mario Sergio Rodriguez

Christian Posse

Matthieu Michel Robin Landos

Rohit Kapoor

Georgios Balikas

Darya Brazouskaya

Qianqian Shi

A database system may receive a natural language query that is associated with a tenant of a multi-tenant system. The natural language query may be parsed into a set of tokens, and the set of tokens may be tagged, using a tenant specific tagging model associated with the tenant, the set of tokens with at least one pre-configured data type identifier that is configured for the plurality of tenants. A global tagging model that supports the plurality of tenants of the multi-tenant system may tag the set of tokens with at least one category identifier. The global tagging model may use the natural language query and the pre-configured data type identifier to identify the at least one category identifier. The system may execute a database query on a database associated with the tenant using the at least one pre-configured data type identifier and the at least one category identifier.

Tenant specific and global pretagging for natural language queries

Methods, systems, and devices supporting detecting and processing conceptual queries are described. A device (e.g., an application server) may receive a search query from a user device. The search query may include one or more parameters. The device may tag the search query using one or more tags associated with the one or more parameters. In some examples, the one or more tags may be determined based on a neural network. The device may determine that the search query is supported as a conceptual query based on a tag of the one or more tags corresponding to a data object stored in a database. The device may then generate a database query in a query language based on the search query, retrieve a set of results for the search query using the database query in the query language, and transmit the set of results to the user device.

Detecting and processing conceptual queries

In disclosed techniques, a computing system causes presentation of a user interface having an input field operable to receive, from a user, a search query for a database. The computing system may classify the search query by: determining whether the search query includes terms that are within a specified vocabulary indicative of a natural language query and determining whether the search query includes terms that identify an object defined in a schema of the database. In response to classifying the search query as a natural language query, the computing system returns query results determined by identifying values in the database corresponding to the object defined in the schema. In response to classifying the search query as a keyword query, the computing system returns query results determined by comparing terms of the search query to values within records in the database.

Classifying different query types

As part of providing the services to users, an online system stores multiple records that are accessible by users of the online system. When a user provides a search query, the online system extracts morphological and dictionary features from the query. The online system provides the extracted features to a machine learning model as an input. The machine learning model outputs a score for each potential entity type that indicates a likelihood that the search query is for a record associated with the entity type. The output from the machine learning model is used by the online system to select one or more entity types that the user is likely searching for. The online system searches the stored records based on the search query but limits the searching to records associated with at least one of the selected entity types.

Predicting a type of a record searched for by a user

An online system stores documents for access by users. The online system also stores query independent information about the documents. Query independent features include data that can be used to score or rank a document independent of any terms entered as a search query. The online system periodically determines whether the values of query independent features have changed, such as by checking activity logs. The online system updates records of query independent features accordingly, and sends information about the updated records to an enterprise search platform for re-indexing. When a user sends a search query to the online system, the enterprise search platform determines whether documents are relevant to the query based on the document contents and the query independent features associated with the documents.

Re-indexing query-independent document features for processing search queries

An online system identifies and ranks records using multiple machine learning models in response to a search query. Therefore, the online system can provide selected records that are of the most relevance to a user of a client device that provided the search query. More specifically, the online system applies a first machine learning model that is of low complexity, such as a regression model. Therefore, the first machine learning model can quickly narrow down the large number of records of the online system to a first set of candidate records. The online system analyzes candidate records in the first set by applying a more complex, second machine learning model that more accurately determines records of interest for the user. In various embodiments, the online system can apply subsequent machine learning models of higher complexity for selecting and ranking records for provision to the client device.

Ranking search results using machine learning based models

A method for configuring a multi-path index includes receiving and storing, in a database management system configured to store a structured document in its native format, a multi-path index definition associated with a data model corresponding to the structured document. In an embodiment, the multi-path index definition includes a sub-path definition that covers a plurality of descendant elements of a root element of the data model and includes at least one index property. Each of a plurality of descendant elements covered by the sub-path definition is automatically indexed according to the at least one index property. The multi-path index definition is stored in a data structure associated with a multi-path index configured to store indexed data from the structured document.

Method and system for configuring a multi-path index

A method for indexing a structured document includes providing a multi-path index definition associated with a data model corresponding to a structured document. The multi-path index definition includes a sub-path definition that covers a root element's descendant elements and includes at least one index property. When a first path expression representing a first descendant element from a first structured document is received, the method includes determining that the first descendant element is covered by the sub-path definition based on the first path expression, indexing the first descendant element according to the index property to generate a path-value pair, and storing the path-value pair and a reference to the first structured document in an inverted multi-path index.

Method and system for indexing a structured document

A method for processing a query includes providing an inverted multi-path index for storing path-value pairs. Each path-value pair references at least one structured document stored in a database system, and comprises an index path expression of an indexed element and an indexed value associated with the indexed element. The method includes receiving a clause including a path expression-value pair comprising a path expression associated with an element, determining that the clause can be processed by the inverted multi-path index, processing the clause to identify a path-value pair in the inverted multi-path index matching the path expression-value pair of the clause, and identifying the structured document referenced by the matching path-value pair.

Growing community of inventors

Voorburg, Netherlands

Francisco Borges