Update README.md

README.md

@@ -12,7 +12,7 @@ widget:
 
 # What is this?
 
-A pretrained BERT model (base version, ~110 M parameters) for Danish NLP. The model was not pre-trained from scratch but adapted from the English version.
+A pre-trained BERT model (base version, ~110 M parameters) for Danish NLP. The model was not pre-trained from scratch but adapted from the English version with a tokenizer trained on Danish text.
 
 # How to use
 
@@ -47,6 +47,6 @@ Initially, only the word token embeddings are trained using 1.000.000 samples. F
 
 # Evaluation
 
-Performance of pretrained model was evaluated using [ScandEval](https://github.com/ScandEval/ScandEval).
+The performance of the pretrained model was evaluated using [ScandEval](https://github.com/ScandEval/ScandEval).
 
-RESULTS
+'| task                     | dataset      | summary                                                                                    |\n|:-------------------------|:-------------|:-------------------------------------------------------------------------------------------|\n| sentiment-classification | swerec       | mcc = 63.02, mcc_se = 2.16, macro_f1 = 62.2, macro_f1_se = 3.61                            |\n| sentiment-classification | angry-tweets | mcc = 47.21, mcc_se = 0.53, macro_f1 = 64.21, macro_f1_se = 0.53                           |\n| sentiment-classification | norec        | mcc = 42.23, mcc_se = 8.69, macro_f1 = 57.24, macro_f1_se = 7.67                           |\n| named-entity-recognition | suc3         | micro_f1 = 50.03, micro_f1_se = 4.16, micro_f1_no_misc = 53.55, micro_f1_no_misc_se = 4.57 |\n| named-entity-recognition | dane         | micro_f1 = 76.44, micro_f1_se = 1.36, micro_f1_no_misc = 80.61, micro_f1_no_misc_se = 1.11 |\n| named-entity-recognition | norne-nb     | micro_f1 = 68.38, micro_f1_se = 1.72, micro_f1_no_misc = 73.08, micro_f1_no_misc_se = 1.66 |\n| named-entity-recognition | norne-nn     | micro_f1 = 60.45, micro_f1_se = 1.71, micro_f1_no_misc = 64.39, micro_f1_no_misc_se = 1.8  |\n| linguistic-acceptability | scala-sv     | mcc = 5.01, mcc_se = 5.41, macro_f1 = 49.46, macro_f1_se = 3.67                            |\n| linguistic-acceptability | scala-da     | mcc = 54.74, mcc_se = 12.22, macro_f1 = 76.25, macro_f1_se = 6.09                          |\n| linguistic-acceptability | scala-nb     | mcc = 19.18, mcc_se = 14.01, macro_f1 = 55.3, macro_f1_se = 8.85                           |\n| linguistic-acceptability | scala-nn     | mcc = 5.72, mcc_se = 5.91, macro_f1 = 49.56, macro_f1_se = 3.73                            |\n| question-answering       | scandiqa-da  | em = 26.36, em_se = 1.17, f1 = 32.41, f1_se = 1.1                                          |\n| question-answering       | scandiqa-no  | em = 26.14, em_se = 1.59, f1 = 32.02, f1_se = 1.59                                         |\n| question-answering       | scandiqa-sv  | em = 26.38, em_se = 1.1, f1 = 32.33, f1_se = 1.05                                          |\n| speed                    | speed        | speed = 4.55, speed_se = 0.0                                                               |'