Title:
MULTI-TYPE ENTITY-BASED KNOWLEDGE COMPLEMENTING METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2021/095987
Kind Code:
A1
Abstract:
Disclosed is a multi-type entity-based knowledge complementing method. The multi-type entity-based knowledge complementing method according to another embodiment of the present invention comprises the steps of: extracting entity information that is information about a plurality of entities included in an individual path and relationship information that is information about a relationship of two of the plurality of entities, for each of a plurality of paths connecting the two entities included in a knowledge graph; generating a path vector corresponding to each of the plurality of paths on the basis of the entity information, the relationship information, type information that is information about at least one entity type corresponding to each of the plurality of entities, and a certain embedding size; calculating an encoded path vector from the path vector by using a Convolutional Neural Network (CNN) and Bidirectional Long A Short-Term Memory (Bi-LSTM); and determining whether a certain target relationship between the two entities is valid by using a relationship model trained to predict a relationship between the two entities by using the plurality of paths and a plurality of encoded path vectors.
Inventors:
PARK YOUNG TACK (KR)
JAGVARAL BATSELEM (KR)
LEE WAN GON (KR)
ROH JAE SEUNG (KR)
JAGVARAL BATSELEM (KR)
LEE WAN GON (KR)
ROH JAE SEUNG (KR)
Application Number:
PCT/KR2019/018701
Publication Date:
May 20, 2021
Filing Date:
December 30, 2019
Export Citation:
Assignee:
FOUNDATION SOONGSIL UNIV INDUSTRY COOPERATION (KR)
International Classes:
G06N5/02; G06F16/901; G06N3/08
Foreign References:
| US20190122111A1 | 2019-04-25 | |||
| KR20190064641A | 2019-06-10 | |||
| CN107832400A | 2018-03-23 | |||
| KR20180092733A | 2018-08-20 |
Other References:
LEE, WAN-GON ET AL.: "Partial Embedding Approach for Knowledge Completion", JOURNAL OF KIISE, November 2018 (2018-11-01), pages 1168 - 1175
Attorney, Agent or Firm:
SHIM, Kyoung-Shik et al. (KR)
Download PDF: