Datasets:
SuzanaB
commited on
Commit
•
440bd9e
1
Parent(s):
6a5ed65
Update readme.md
Browse files
README.md
CHANGED
@@ -21,5 +21,10 @@ The hr500k training corpus contains about 500,000 Croatian tokens manually annot
|
|
21 |
|
22 |
On the sentence level, the dataset contains 20159 training samples, 1963 validation samples and 2672 test samples
|
23 |
across the respective data splits. Each sample represents a sentence and includes the following features:
|
24 |
-
sentence ID ('sent\_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
On the sentence level, the dataset contains 20159 training samples, 1963 validation samples and 2672 test samples
|
23 |
across the respective data splits. Each sample represents a sentence and includes the following features:
|
24 |
+
sentence ID ('sent\_id'), sentence text ('text'), list of tokens ('tokens'), list of lemmas ('lemmas'),
|
25 |
+
list of Multext-East tags ('xpos\_tags), list of UPOS tags ('upos\_tags'), list of morphological features ('feats'),
|
26 |
+
and list of IOB tags ('iob\_tags'). A subset of the data also contains universal dependencies ('ud') and consists of
|
27 |
+
7498 training samples, 649 validation samples, and 742 test samples.
|
28 |
+
|
29 |
+
Three dataset configurations are available, namely 'ner', 'upos', and 'ud', with the corresponding features
|
30 |
+
encoded as class labels. If the configuration is not specified, it defaults to 'ner'.
|