Targeted PMI scores between the arg1 of loss and other argument positions. The second column gives the number of times that the argument position in the row is found to be coreferent with the arg1 of the loss predicate. A higher value in this column results in a lower discount factor. See Equation (4) for the discount factor.
| Argument position . | #coref with loss.arg1 . | Raw PMI score . | Discounted PMI score . |
|---|---|---|---|
| win.arg1 | 37 | 5.68 | 5.52 |
| gain.arg1 | 10 | 5.13 | 4.64 |
| recoup.arg1 | 2 | 6.99 | 4.27 |
| steal.arg1 | 4 | 5.18 | 4.09 |
| possess.arg1 | 3 | 5.10 | 3.77 |
| Argument position . | #coref with loss.arg1 . | Raw PMI score . | Discounted PMI score . |
|---|---|---|---|
| win.arg1 | 37 | 5.68 | 5.52 |
| gain.arg1 | 10 | 5.13 | 4.64 |
| recoup.arg1 | 2 | 6.99 | 4.27 |
| steal.arg1 | 4 | 5.18 | 4.09 |
| possess.arg1 | 3 | 5.10 | 3.77 |