Table 9:

The effect of each term in the scoring function Eq. (3). discriminative stands for $logp(y=1|si,ci)$, left-context stands for $1|ci)$ and right-context stands for $1|c>i|logp(c>i|c. both contexts means we remove both left context and right context.

ModelSpearmanρ
Full 78.47
w/o discriminative 77.97 (−0.50)
w/o left-context 77.36 (−1.11)
w/o right-context 77.01 (−1.46)
w/o both contexts 76.50 (−1.97)
ModelSpearmanρ
Full 78.47
w/o discriminative 77.97 (−0.50)
w/o left-context 77.36 (−1.11)
w/o right-context 77.01 (−1.46)
w/o both contexts 76.50 (−1.97)
Close Modal