US 12,013,884 B2
Knowledge graph question answering with neural machine translation
Saswati Dana, Bangalore (IN); Dinesh Garg, Beawar (IN); Dinesh Khandelwal, Indore (IN); G P Shrivatsa Bhargav, Bengaluru (IN); and Sukannya Purkayastha, Dhubri (IN)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on Jun. 30, 2022, as Appl. No. 17/810,013.
Prior Publication US 2024/0004907 A1, Jan. 4, 2024
Int. Cl. G06F 16/332 (2019.01); G06F 40/30 (2020.01); G06N 3/04 (2023.01); G06N 5/02 (2023.01)
CPC G06F 16/3329 (2019.01) [G06F 40/30 (2020.01); G06N 3/04 (2013.01); G06N 5/02 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving an input question for a question-answering (QA) system;
masking, by a noise simulator, entities and relations of the input question to generate a masked input question;
translating the masked input question to a SPARQL Protocol and RDF Query Language (SPARQL) query silhouette using neural machine translation (NMT);
generating, by a neural network graph search module using the SPARQL query silhouette, a finalized SPARQL query, wherein the neural network graph search module is a Bidirectional Encoder Representations from Transformers (BERT) based graph search module; and
querying, using the finalized SPARQL query, the QA system to obtain an answer to the received input question.