Dynamic example generation for queries
First Claim
Patent Images
1. A method of generating example natural language queries, comprising:
- receiving an actual natural language query from a user;
partitioning the actual natural language query into one or more tokens;
generating one or more expressions in a precise syntax using the one or more tokens;
processing the one or more expressions in an attempt to generate an answer to the natural language query;
generating one or more fragments of the actual natural language query based on reassembly of the one or more tokens, wherein at least one of the one or more fragments has multiple portions comprising the one or more tokens;
generating a plurality of example natural language queries based on (i) different interpretations of the one or more fragments, and (ii) a plurality of query templates for comparing to the one or more fragments, wherein at least some of the plurality of query templates comprise pairs of (i) known categories and (ii) known properties respectively associated with the categories, and wherein generating the plurality of example natural language queries includes;
determining, for each fragment having multiple portions, whether a first portion of the fragment matches a first category corresponding to a first one of the plurality of query templates and a second portion of the fragment matches a first property corresponding to the first one of the plurality of query templates, the first property associated with the first category, andwhen it is determined that the first portion of the fragment matches the first category corresponding to the first one of the plurality of query templates and the second portion of the fragment matches the first property corresponding to the first one of the plurality of query templates, generating one of the plurality of example natural language queries based on the first one of the plurality of query templates; and
transmitting to the user, via a network, the at least one of the plurality of example natural language queries.
2 Assignments
0 Petitions
Accused Products
Abstract
In a method for generating example queries, an actual query is received, and the actual query is portioned into one or more tokens. One or more fragments of the actual query are generated based on reassembly of the one or more tokens. A plurality of example queries are generated based on different interpretations of one or more fragments, and at least one example query is outputted.
-
Citations
18 Claims
-
1. A method of generating example natural language queries, comprising:
-
receiving an actual natural language query from a user; partitioning the actual natural language query into one or more tokens; generating one or more expressions in a precise syntax using the one or more tokens; processing the one or more expressions in an attempt to generate an answer to the natural language query; generating one or more fragments of the actual natural language query based on reassembly of the one or more tokens, wherein at least one of the one or more fragments has multiple portions comprising the one or more tokens; generating a plurality of example natural language queries based on (i) different interpretations of the one or more fragments, and (ii) a plurality of query templates for comparing to the one or more fragments, wherein at least some of the plurality of query templates comprise pairs of (i) known categories and (ii) known properties respectively associated with the categories, and wherein generating the plurality of example natural language queries includes; determining, for each fragment having multiple portions, whether a first portion of the fragment matches a first category corresponding to a first one of the plurality of query templates and a second portion of the fragment matches a first property corresponding to the first one of the plurality of query templates, the first property associated with the first category, and when it is determined that the first portion of the fragment matches the first category corresponding to the first one of the plurality of query templates and the second portion of the fragment matches the first property corresponding to the first one of the plurality of query templates, generating one of the plurality of example natural language queries based on the first one of the plurality of query templates; and transmitting to the user, via a network, the at least one of the plurality of example natural language queries. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
-
-
9. A system for generating example natural language queries, the system comprising a plurality of modules embodied as blocks in a hardware processor, the modules comprising:
-
a parser module to generate one or more tokens from an actual natural language query received from a user; a plurality of fragment generator modules to generate at least one fragment of the actual natural language query based on reassembly of the one or more tokens, wherein at least one of the one or more fragments has multiple portions comprising the one or more tokens; and a plurality of example natural language query generator modules to generate a plurality of example natural language queries based on (i) different interpretations of the at least one fragment, and (ii) a plurality of query templates for comparing to the at least one fragment, the query templates retrieved from a database, wherein at least some of the plurality of query templates comprise pairs of (i) known categories and (ii) known properties respectively associated with the categories, and wherein the plurality of example natural language query generator modules generate the plurality of example natural language queries by; determining, for each fragment having multiple portions, whether a first portion of the fragment matches a first category corresponding to a first one of the example query templates and a second portion of the fragment matches a first property corresponding to the first one of the example query templates, the first property associated with the first category, and when it is determined that the first portion of the fragment matches the first category corresponding to the first one of the plurality of query templates and the second portion of the fragment matches the first property corresponding to the first one of the plurality of query templates, generating one of the plurality of example natural language queries based on the first one of the plurality of query templates. - View Dependent Claims (10, 11, 12, 13, 14, 15)
-
-
16. A method, comprising:
-
receiving from a user a query in an imprecise syntax; transmitting, the query in the imprecise syntax; receiving at least one user interface mechanism to permit selection of one of at least one example natural language query, the at least one example natural language query having been dynamically generated in response to the query in the imprecise syntax, based on (i) different interpretations of one or more fragments generated by reassembling tokens into which the query in the imprecise syntax has been partitioned and (ii) a plurality of query templates compared to the one or more fragments, wherein at least one of the one or more fragments has multiple portions comprising the tokens, wherein (a) at least some of the plurality of query templates comprise pairs of (i) known categories and (ii) known properties respectively associated with the categories and (b) the at least one example natural language query was generated in response to a determination that a first portion of a fragment matched a category of a first one of the plurality of query templates and a second portion of the fragment matched a property corresponding to the first one of the plurality of query templates; receiving a user input that corresponds to a selection of one of the at least one example natural language query; transmitting an indication of the selected example natural language query; and receiving query results in response to the selected example natural language query. - View Dependent Claims (17, 18)
-
Specification