Natural language querying of data in a structured context
AssigneeSparkcognition, Inc.
InventorsSyed Mohammad Ali, Erik Skiles
Life of this patent
Estimating the term…
Abstract
A particular method includes automatically generating, at a processor of a computing device, annotation data indicating that a column of a data table corresponds to a particular class of an ontology. The method also includes storing the annotation data. The method further includes receiving a natural language query. The method also includes generating a second query based on detecting a match between at least one term of the natural language query and the annotation data. The method further includes determining a response to the second query. The method also includes outputting the response to the second query as a response to the natural language query.
Claims23
A method of querying a data table, the method comprising:
- receiving a natural language query;
- providing a vector representing the natural language query as input to a trained data model;
- determining, based on output of the trained data model, similarity metrics indicative of proximity of the vector to points in a feature space, the similarity metrics comprising a first similarity metric based on a comparison of the vector with a first ontology word vector and a second similarity metric based on a comparison of the vector with a second ontology word vector;
- selecting a subset of the data table based on the similarity metrics;
- searching the subset of the data table to determine a response to the natural language query;
- and
- outputting the response to the natural language query.
A computing device comprising:
- a memory storing a data table;
- an input interface configured to receive a natural language query;
- a processor configured to:
- provide a vector representing the natural language query as input to a trained data model;
- determine, based on output of the trained data model, similarity metrics indicative of proximity of the vector to points in a feature space, the similarity metrics comprising a first similarity metric based on a comparison of the vector with a first ontology word vector and a second similarity metric based on a comparison of the vector with a second ontology word vector;
- select a subset of the data table based on the similarity metrics;
- and
- search the subset of the data table to determine a response to the natural language query;
- and
- an output interface configured to output the response to the natural language query.
A computer-readable storage device storing instructions that, when executed, causes a processor to perform operations comprising:
- receiving a natural language query;
- providing a vector representing the natural language query as input to a trained data model;
- determine, based on output of the trained data model, similarity metrics indicative of proximity of the vector to points in a feature space, the similarity metrics comprising a first similarity metric based on a comparison of the vector with a first ontology word vector and a second similarity metric based on a comparison of the vector with a second ontology word vector;
- selecting a subset of a data table based on the similarity metrics;
- searching the subset of the data table to determine a response to the natural language query;
- and
- outputting the response to the natural language query.
Dependencies are inferred from the claim text.