Datasets:
Clarify SINS common-ground judgment terminology
Browse files
README.md
CHANGED
|
@@ -17,11 +17,11 @@ size_categories:
|
|
| 17 |
|
| 18 |
## Dataset Description
|
| 19 |
|
| 20 |
-
SINS is a 13,077-instance binary common-ground judgment
|
| 21 |
-
|
| 22 |
-
MapTask (GMMT) annotations. Each instance asks whether the giver and
|
| 23 |
-
interpret the referring expression as the same landmark. Rows
|
| 24 |
-
identifiers and context-window provenance while mapping GMMT status to
|
| 25 |
`gold_label`: `aligned` becomes `yes`; `pending` and `misunderstood` become
|
| 26 |
`no`.
|
| 27 |
|
|
|
|
| 17 |
|
| 18 |
## Dataset Description
|
| 19 |
|
| 20 |
+
SINS is a 13,077-instance binary common-ground judgment (interpretation
|
| 21 |
+
matching judgment) dataset, derived from the public Grounded Misunderstandings
|
| 22 |
+
in MapTask (GMMT) annotations. Each instance asks whether the giver and
|
| 23 |
+
follower interpret the referring expression as the same landmark. Rows
|
| 24 |
+
preserve identifiers and context-window provenance while mapping GMMT status to
|
| 25 |
`gold_label`: `aligned` becomes `yes`; `pending` and `misunderstood` become
|
| 26 |
`no`.
|
| 27 |
|