Update README

README.md

@@ -40,13 +40,34 @@ tags:
 - **Point of Contact:** TBA
 
 ### Dataset Summary
-This is the oficial repository for Super TweetEval
+This is the oficial repository for SuperTweetEval, a unified benchmark of 12 heterogeneous NLP tasks. 
+More details on the task and an evaluation of language models can be found on the reference paper.
 
 
+### Data Splits
+
+All tasks provide custom training, validation and test splits.
+
+| task             | description                        | number of instances   |
+|:-----------------|:-----------------------------------|:----------------------|
+| tweet_topic      | multi-label classification         | 4,585 / 573 / 1,679     |
+| tweet_ner7       | sequence labeling                  | 4,616 / 576 / 2,807     |
+| tweet_qa         | generation                         | 9,489 / 1,086 / 1,203    |
+| tweet_qg         | generation                         | 9,489 / 1,086 / 1,203    |
+| tweet_intimacy   | regression on a single text        | 1,191 / 396 / 396      |
+| tweet_similarity | regression on two texts            | 450 / 100 / 450       |
+| tempo_wic        | binary classification on two texts | 1,427 / 395 / 1,472     |
+| tweet_hate       | multi-class classification         | 5,019 / 716 / 1,433     |
+| tweet_emoji      | multi-class classification         | 50,000 / 5,000 / 50,000 |
+| tweet_sentiment  | ABSA on a five-pointscale          | 26,632 / 4,000 / 12,379  |
+| tweet_nerd       | binary classification              | 20,164 / 4,100 / 20,075           |
+| tweet_emotion    | multi-label classification      | 6,838 / 886 / 3,259      |
+
 ## Dataset Structure
 ### Data Fields
 
-The data fields are the same among all splits.
+The data fields are unified among all splits. 
+In the following we present the information contained in each of the datasets.
 
 #### tweet_topic
 - `text`: a `string` feature.
@@ -118,23 +139,6 @@ The data fields are the same among all splits.
 - `gold_label_list`: a list of  `string` feature.
 
 
-### Data Splits
-
-| task             | description                        | number of instances   |
-|:-----------------|:-----------------------------------|:----------------------|
-| tweet_topic      | multi-label classification         | 4,585 / 573 / 1,679     |
-| tweet_ner7       | sequence labeling                  | 4,616 / 576 / 2,807     |
-| tweet_qa         | generation                         | 9,489 / 1,086 / 1,203    |
-| tweet_qg         | generation                         | 9,489 / 1,086 / 1,203    |
-| tweet_intimacy   | regression on a single text        | 1,191 / 396 / 396      |
-| tweet_similarity | regression on two texts            | 450 / 100 / 450       |
-| tempo_wic        | binary classification on two texts | 1,427 / 395 / 1,472     |
-| tweet_hate       | multi-class classification         | 5,019 / 716 / 1,433     |
-| tweet_emoji      | multi-class classification         | 50,000 / 5,000 / 50,000 |
-| tweet_sentiment  | ABSA on a five-pointscale          | 26,632 / 4,000 / 12,379  |
-| tweet_nerd       | binary classification              | 20,164 / 4,100 / 20,075           |
-| tweet_emotion    | multi-label classification      | 6,838 / 886 / 3,259      |
-
 
 ## Evaluation Metrics
 - __tweet_topic:__ ```macro-F1```
@@ -161,6 +165,19 @@ The data fields are the same among all splits.
 
 ## Citation Information
 
+### Main reference paper
+
+Please cite the [reference paper]() if you use this benchmark.
+
+```bibtex
+TBA
+```
+
+### References of individual datasets
+
+In addition to the main reference paper, please cite the individual task datasets included in SuperTweetEval if you use them.
+
+
 - TweetTopic
 ```
 @inproceedings{antypas-etal-2022-twitter,