当前位置:网站首页>TEC: Knowledge Graph Embedding with Triple Context

TEC: Knowledge Graph Embedding with Triple Context

2022-07-01 03:33:00 Re:fused

TEC:Knowledge Graph Embedding with Triple Context

1 problem

    The knowledge map is not fully utilized , Rich information , Use its surrounding information and path information , Complete the knowledge map based on the above information .

2 Model

context The information includes :neighbor context and path context.
 Insert picture description here

2.1 Context Details

  • Neighbor Context
    For its Neighbor Context Only calculate the output , C N ( h ) = ( r 4 , e 1 ) , ( r 3 , e 2 ) , ( r 2 , e 3 ) , ( r 1 , e 8 ) , ( r 1 , e 1 0 ) CN (h) = {(r_4, e_1), (r_3, e_2), (r_2, e_3), (r_1, e_8), (r_1, e_10)} CN(h)=(r4,e1),(r3,e2),(r2,e3),(r1,e8),(r1,e10)
  • Path Context
    about Path Context, C P ( h , t ) = ( r 1 , r 2 ) , ( r 2 , r 1 , r 2 ) . CP (h,t) = {(r_1,r_2), (r_2,r_1,r_2)}. CP(h,t)=(r1,r2),(r2,r1,r2)., Save only h and t The relationship between .
    C ( h , r , t ) = C N ( h ) ∪ C P ( h , t ) C(h,r,t) = CN (h) ∪CP (h,t) C(h,r,t)=CN(h)CP(h,t)

2.2 scoring function

 Insert picture description here
Maximize possibilities :
 Insert picture description here
It can be written in the following form :
 Insert picture description here
 Insert picture description here
 Insert picture description here
 Insert picture description here
 Insert picture description here
 Insert picture description here
 Insert picture description here
 Insert picture description here
Train through the above formula , Consumption of resources , Therefore, negative sampling is used for training
 Insert picture description here
The other two are similar .

原网站

版权声明
本文为[Re:fused]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/182/202207010316404936.html