Title:
ENTITY NAME CORRECTION SYSTEM AND METHOD FOR TRAINING SAME
Document Type and Number:
WIPO Patent Application WO/2023/085506
Kind Code:
A1
Abstract:
Provided is an entity name correction system. The entity name correction system may comprise: an entity name extraction component configured to acquire original text entity names from an original text and to acquire abstract entity names from an abstract; a language component configured to acquire original text entity name vectors for the original text entity names from original text sentences corresponding to the original text entity names and to acquire abstract entity name vectors for the abstract entity names from abstract sentences corresponding to the abstract entity names; a determination component configured to determine, from a target abstract entity name vector among the abstract entity name vectors and the original text entity name vectors, whether a target abstract entity name corresponding to the target abstract entity name vector is correct in light of the original text; and a correction component configured to change, if the target abstract entity name is incorrect, the target abstract entity name into one of the original text entity names or a new entity name different from the original text entity names, wherein the determination component and the correction component share a common subcomponent configured to acquire encoder hidden states, decoder hidden states, attention distribution, and a context vector from the original text entity name vectors and the target abstract entity name vector.
Inventors:
LEE KYUNG IL (KR)
PARK JUN MO (KR)
PARK JUN MO (KR)
Application Number:
PCT/KR2021/018735
Publication Date:
May 19, 2023
Filing Date:
December 10, 2021
Export Citation:
Assignee:
SALTLUX INC (KR)
International Classes:
G06F40/295; G06F40/232; G06N3/04; G06N3/08
Foreign References:
| EP0361464A2 | 1990-04-04 | |||
| KR101713831B1 | 2017-03-09 |
Other References:
JEONGWAN SHIN, YUNSEOK NOH, SANGHEON PARK, YOUNGSUN O, SEYOUNG PARK: "Multi-task learning for entity-centric fact correction on machine summaries", PROCEEDINGS OF THE ANNUAL CONFERENCE ON HUMAN & COGNITIVE LANGUAGE TECHNOLOGY, KR, no. 10a, 14 October 2021 (2021-10-14), KR , pages 124 - 130, XP009546982, ISSN: 2005-3053
GYOUNG HO LEE, KONG JOO LEE: "Single Document Extractive Summarization Based on Deep Neural Networks Using Linguistic Analysis Features).", KIPS TRANSACTIONS ON SOFTWARE AND DATA ENGINEERING, vol. 8, no. 8, 19 May 2019 (2019-05-19), pages 343 - 348, XP093066415, ISSN: 2287-5905, DOI: 10.3745/KTSDE.2019.8.8.343
LEE DONGYUB, SHIN MYEONG CHEOL, WHANG TAESUN, CHO SEUNGWOO, KO BYEONGIL, LEE DANIEL, KIM EUNGGYUN, JO JAECHOON: "Reference and Document Aware Semantic Evaluation Methods for Korean Language Summarization", PROCEEDINGS OF THE 28TH INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS, INTERNATIONAL COMMITTEE ON COMPUTATIONAL LINGUISTICS, STROUDSBURG, PA, USA, 1 January 2020 (2020-01-01), Stroudsburg, PA, USA, pages 5604 - 5616, XP093066416, DOI: 10.18653/v1/2020.coling-main.491
GYOUNG HO LEE, KONG JOO LEE: "Single Document Extractive Summarization Based on Deep Neural Networks Using Linguistic Analysis Features).", KIPS TRANSACTIONS ON SOFTWARE AND DATA ENGINEERING, vol. 8, no. 8, 19 May 2019 (2019-05-19), pages 343 - 348, XP093066415, ISSN: 2287-5905, DOI: 10.3745/KTSDE.2019.8.8.343
LEE DONGYUB, SHIN MYEONG CHEOL, WHANG TAESUN, CHO SEUNGWOO, KO BYEONGIL, LEE DANIEL, KIM EUNGGYUN, JO JAECHOON: "Reference and Document Aware Semantic Evaluation Methods for Korean Language Summarization", PROCEEDINGS OF THE 28TH INTERNATIONAL CONFERENCE ON COMPUTATIONAL LINGUISTICS, INTERNATIONAL COMMITTEE ON COMPUTATIONAL LINGUISTICS, STROUDSBURG, PA, USA, 1 January 2020 (2020-01-01), Stroudsburg, PA, USA, pages 5604 - 5616, XP093066416, DOI: 10.18653/v1/2020.coling-main.491
Attorney, Agent or Firm:
Y.P.LEE, MOCK & PARTNERS (KR)
Download PDF: