US 11,741,161 B2
System and method for generating a refined query
Thomas Hubauer, Garching bei München (DE); Swathi Shyam Sunder, Munich (DE); and Janaki Joshi, Karnataka (IN)
Assigned to SIEMENS AKTIENGESELLSCHAFT
Filed by Siemens Aktiengesellschaft, Munich (DE)
Filed on May 5, 2021, as Appl. No. 17/308,377.
Prior Publication US 2022/0358166 A1, Nov. 10, 2022
Int. Cl. G06F 16/90 (2019.01); G06F 16/9032 (2019.01); G06F 16/901 (2019.01); G06F 16/953 (2019.01); G06F 11/34 (2006.01); G06F 16/9035 (2019.01); G06F 16/20 (2019.01); G06F 11/30 (2006.01)
CPC G06F 16/90324 (2019.01) [G06F 11/3409 (2013.01); G06F 16/9027 (2019.01); G06F 16/9035 (2019.01); G06F 16/953 (2019.01)] 11 Claims
OG exemplary drawing
 
1. A system for automatically generating a refined query for a currently given query, wherein the system comprises at least one processor and further comprises a search engine for searching through a tree of query modification operations, wherein a root node of said tree is an empty node which represents a given initial query,
wherein the at least one processor is configured to perform the following steps:
a) defining a set of query modification operators which can be inserted into said tree;
b) receiving a first set of current query results from a currently given query comprising one or more triple patterns;
c) receiving a second set of reference query results, wherein the second set of reference query results are determined by user selection of at least a subset of the first set of current query results and/or addition of at least a set of missing query results into the first set of current query results;
d) contrasting the first set of current query results with the second set of reference query results by assessing differences between the first set of current query results and the second set of reference query results, wherein assessing differences includes deriving a computed score including a number of unwanted answers and a number of missing answers; and
wherein the search engine is configured to perform the following steps:
e) selecting a node of said tree by the computed score derived from the assessed differences;
f) selecting any query modification operator of the defined set of query modification operators;
g) if the selected query modification operator does not correspond to any triple patter of the query represented by the selected node, then returning to step e), otherwise continuing to step h;
h) identifying at least one triple pattern of the query which the selected query modification operator corresponds to; and
i) automatically generating the refined query by applying the selected query modification operator to the identified at least one triple pattern;
wherein the search engine interacts with a machine learning model that learns from the currently given query, from the first set of current query results, from the second set of reference query results as well as from the assessed differences between these two query results, from the selected node and from the selected query modification operator in order to automatically generate the refined query.