search

NUSTM / ECPE

165 stars 43 forks source link

Distance for filtering emotion cause pairs #10

Open bhshri opened 4 years ago

bhshri commented 4 years ago

The paper mentions distance to be used in filtering the candidate emotion cause pairs. What is this distance that has been used ? Screenshot_2020-04-09-00-48-12-448_com google android apps docs

bbruceyuan commented 4 years ago

In this paper, the second step is to filter the no causal emotion-cause pairs using a logistic regression model. And the inputs of this model is an emotion clause and cause clause and the distance embedding. Here, the distance between them is emotion clause index i minus cause clause j. Then get the distance embedding through a position lookup table.

Hoping it solves your problem!