mathiascreutz
commited on
Commit
·
fa517c7
1
Parent(s):
29a456d
Minor modifications
Browse files
README.md
CHANGED
@@ -240,13 +240,27 @@ Copenhagen, Denmark.
|
|
240 |
|
241 |
### Data Splits
|
242 |
|
243 |
-
The data is split into
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
250 |
|
251 |
|
252 |
## Dataset Creation
|
|
|
240 |
|
241 |
### Data Splits
|
242 |
|
243 |
+
The data is split into training, validation and test sets. The
|
244 |
+
validation and test sets come in two versions, the regular validation
|
245 |
+
and test sets and the full sets validation.full and test.full. The
|
246 |
+
full sets contain all sentence pairs successfully annotated by the
|
247 |
+
annotators, including the sentence pairs that were rejected as
|
248 |
+
paraphrases. The annotation score of the full sets thus range between
|
249 |
+
1.0 and 4.0. The regular validation and test sets only contain
|
250 |
+
paraphrases, scored between 3.0 and 4.0 by the annotators.
|
251 |
+
|
252 |
+
The number of sentence pairs in the data splits are as follows for
|
253 |
+
each of the languages. The range between the smallest (`quality=95`)
|
254 |
+
and largest (`quality=60`) train configuration have been shown.
|
255 |
+
|
256 |
+
| | Train | Valid | Test | Valid.full | Test.full |
|
257 |
+
| ----- | ------ | ----- | ---- | ---------- | --------- |
|
258 |
+
| de | 0.59M .. 13M | 1013 | 1047 | 1582 | 1586 |
|
259 |
+
| en | 1.0M .. 35M | ... | ... |
|
260 |
+
| fi | 0.48M .. 8.9M | ... | ... |
|
261 |
+
| fr | 0.94M .. 22M | ... | ... |
|
262 |
+
| ru | 0.15M .. 15M | ... | ... |
|
263 |
+
| sv | 0.24M .. 4.5M | ... | ... |
|
264 |
|
265 |
|
266 |
## Dataset Creation
|