Each word also has a value: what it says. Here, cat says “tired”, mat says “soft”. Write attend(query, keys, values): what “it” collects by listening to every word with the shares from the last puzzle. Keys are one number per word this time (how much of an animal), so the shares are just the chances of keys @ query.
| animal |
|---|
| 1 |
| animal | |
|---|---|
| cat | 2 |
| sat | 0 |
| mat | −1 |
| tired | soft | |
|---|---|---|
| cat | 1 | 0 |
| sat | 0 | 0 |
| mat | 0 | 1 |