Update README.md
Browse files
README.md
CHANGED
@@ -90,10 +90,18 @@ EUROPA is a dataset designed for training and evaluating multilingual keyphrase
|
|
90 |
|
91 |
## Dataset Structure
|
92 |
|
93 |
-
- **celex_id:** CELEX identifier inherited from CJEU. Different translated versions of the same judgment share the same celex_id.
|
94 |
-
- **lang:** ISO 639-1 language code
|
95 |
-
- **input:** judgment transcription or translation
|
96 |
-
- **keyphrases:** reference keyphrases drafted by the CJEU.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
97 |
|
98 |
## Citation
|
99 |
|
|
|
90 |
|
91 |
## Dataset Structure
|
92 |
|
93 |
+
- **celex_id:** CELEX identifier inherited from CJEU. Different translated versions of the same judgment share the same celex_id. If you wish to have a unique identifier for each instance, you can concatenate `lang` and `celex_id` values;
|
94 |
+
- **lang:** ISO 639-1 language code;
|
95 |
+
- **input:** judgment transcription or translation;
|
96 |
+
- **keyphrases:** reference keyphrases drafted by the CJEU.
|
97 |
+
|
98 |
+
This page presents an randomly split version of the dataset, thus allowing all sets to have the same distribution in terms of vocabulary and languages. More details can be found in Appendix H of our paper.
|
99 |
+
|
100 |
+
- **training set**: 159 306 instances;
|
101 |
+
- **validation set**: 53 943 instances;
|
102 |
+
- **test set**: 71 708 instances.
|
103 |
+
|
104 |
+
|
105 |
|
106 |
## Citation
|
107 |
|