sjrhuschlee
commited on
Commit
•
7cd7dcc
1
Parent(s):
96a4f20
Update README.md
Browse files
README.md
CHANGED
@@ -38,7 +38,7 @@ embeds_dropout_prob = 0.1
|
|
38 |
We evaluated the extractive question answering performance on our GermanQuAD test set.
|
39 |
Model types and training data are included in the model name.
|
40 |
For finetuning XLM-Roberta, we use the English SQuAD v2.0 dataset.
|
41 |
-
The GELECTRA models are warm started on the German translation of SQuAD v1.1 and finetuned on
|
42 |
The human baseline was computed for the 3-way test set by taking one answer as prediction and the other two as ground truth.
|
43 |
![performancetable](https://images.prismic.io/deepset/1c63afd8-40e6-4fd9-85c4-0dbb81996183_german-qa-vs-xlm-r.png)
|
44 |
|
38 |
We evaluated the extractive question answering performance on our GermanQuAD test set.
|
39 |
Model types and training data are included in the model name.
|
40 |
For finetuning XLM-Roberta, we use the English SQuAD v2.0 dataset.
|
41 |
+
The GELECTRA models are warm started on the German translation of SQuAD v1.1 and finetuned on [GermanQuAD](https://deepset.ai/germanquad).
|
42 |
The human baseline was computed for the 3-way test set by taking one answer as prediction and the other two as ground truth.
|
43 |
![performancetable](https://images.prismic.io/deepset/1c63afd8-40e6-4fd9-85c4-0dbb81996183_german-qa-vs-xlm-r.png)
|
44 |
|