Title:
KNOWLEDGE GRAPH CONSTRUCTION METHOD FOR EVENT PREDICTION AND EVENT PREDICTION METHOD
Document Type and Number:
WIPO Patent Application WO/2020/232943
Kind Code:
A1
Abstract:
Disclosed are a knowledge graph construction method for event prediction and an event prediction method. The knowledge graph construction method comprises: preprocessing a pre-collected corpus, and extracting a plurality of candidate sentences from the corpus; extracting a plurality of events from the candidate sentences according to a preset dependency relationship, so that each of the events retains complete semantic information corresponding to the candidate sentence; extracting a seed relationship between the events from the corpus; performing possibility relation extraction on the events by means of a pre-constructed relation self-recommendation network model according to the events and the seed relations between the events to obtain candidate event relations between the events; and generating a knowledge graph of events according to the events and the candidate event relations between the events. Common grammatical modes are extracted according to the dependency relations, so as to extract the events with complete semantics from corpora, the activities, states, events and relations between the events can be effectively mined, to construct a high-quality and effective knowledge graph.
Inventors:
ZHANG HONGMING (CN)
LIU XIN (CN)
PAN HAOJIE (CN)
SONG YANGQIU (CN)
LIU XIN (CN)
PAN HAOJIE (CN)
SONG YANGQIU (CN)
Application Number:
PCT/CN2019/108129
Publication Date:
November 26, 2020
Filing Date:
September 26, 2019
Export Citation:
Assignee:
GUANGZHOU HKUST FOK YING TUNG RES INSTITUTE (CN)
International Classes:
G06F16/36; G06F16/332
Foreign References:
| CN110263177A | 2019-09-20 | |||
| CN107358315A | 2017-11-17 | |||
| CN107656921A | 2018-02-02 | |||
| CN103999081A | 2014-08-20 | |||
| US20150127323A1 | 2015-05-07 | |||
| CN109446341A | 2019-03-08 |
Attorney, Agent or Firm:
SCIHEAD IP LAW FIRM (CN)
Download PDF: