The Design And Implementation Of An N-ary Relation Extraction System Based On Knowledge Graphs

Relation Extraction is an important research topic of information retrieval and natural lan-guage processing.Researches on binary relation extraction have been maturing over the years while n-ary relation extraction is still a relatively new field,despite the fact that it is very com-mon for n-ary relations to be mentioned in texts.Knowledge graphs are another place where n-ary relations widely exists.Consisting of multiple binary sub-relations,they exist in a more structured way compared to those in texts.There are few works on discovering n-ary relations in knowledge graphs.This thesis proposes an approach for mining n-ary relations out of knowledge graphs,and extract them on texts with the mined results and existing researches.The primary study in this thesis is described as below:1.Proposing an n-ary relation mining algorithm that discovers n-ary relations in knowledge graphs by analysing their structural characteristics.2.Generating samples of n-nary relations in text form by utilizing n-ary relations in knowl-edge graphs and distant supervision,and then training models with the samples for n-ary relation extraction on texts.3.Designing and implementing an n-ary relation extraction system based on the methods above.In this thesis,n-ary relations in knowledge graphs are connected with those in texts.This thesis presents an application scenario for n-ary relations in knowledge graphs and an inspiration about generating text samples of n-ary relations.