About vector "left" , "right" and function “broadcasting”

[yxWisdom]

Hi， I don't quite understand the use of the vector "left" and "right" and the function “broadcasting”, could you explain it more clearly？ thanks

[sanexodus]

Hi， I don't quite understand the use of the vector "left" and "right" and the function “broadcasting”, could you explain it more clearly？ thanks

Hi yxWisdom, Do you have any idea about that question? Thank you.

[gao929165733]

I also want to know that

[disanrenchengcc]

oriLeft and oriRight shape: [B,T,U]
left transpose [1,0,2]: [T,B,U]；than expand dim 3: [T,B,U,1]
right transpose [0,2,1]:[B,U,T]；than expand dim 0: [1,B,U,T]
B = left + right: [T,B,U,T]
B transpose [1,0,3,2] : [B,T,T,U]
This means that in each sample, each word belongs to the confidence level of each relationship with other words