initial commit

README.md

@@ -0,0 +1,43 @@
+---
+language:
+- da
+tags:
+- bert
+- pytorch
+- hatespeech
+license: cc-by-sa-4.0
+datasets:
+- social media
+metrics:
+- f1
+widget:
+- text: "Senile gamle idiot"
+---
+
+# Danish BERT for hate speech classification
+
+The BERT HateSpeech model classifies offensive Danish text into 4 categories: 
+ * `Særlig opmærksomhed` (special attention, e.g. threat)
+ * `Personangreb` (personal attack) 
+ * `Sprogbrug` (offensive language)
+ * `Spam & indhold` (spam)
+This model is intended to be used after the [BERT HateSpeech detection model](https://huggingface.co/DaNLP/da-bert-hatespeech-detection).
+
+It is based on the pretrained [Danish BERT](https://github.com/certainlyio/nordic_bert) model by BotXO which has been fine-tuned on social media data. 
+
+See the [DaNLP documentation](https://danlp-alexandra.readthedocs.io/en/latest/docs/tasks/hatespeech.html#bertdr) for more details. 
+
+
+Here is how to use the model:
+
+```python
+from transformers import BertTokenizer, BertForSequenceClassification
+
+model = BertForSequenceClassification.from_pretrained("DaNLP/da-bert-hatespeech-classification")
+tokenizer = BertTokenizer.from_pretrained("DaNLP/da-bert-hatespeech-classification")
+```
+
+## Training data
+
+The data used for training has not been made publicly available. It consists of social media data manually annotated in collaboration with Danmarks Radio.
+

config.json

@@ -0,0 +1,41 @@
+{
+  "_name_or_path": ".",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "directionality": "bidi",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "S\u00e6rlig opm\u00e6rksomhed",
+    "1": "Personangreb",
+    "2": "Sprogbrug",
+    "3": "Spam & indhold"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "Personangreb": 1,
+    "Spam & indhold": 3,
+    "Sprogbrug": 2,
+    "S\u00e6rlig opm\u00e6rksomhed": 0
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.5.0",
+  "type_vocab_size": 2,
+  "vocab_size": 32000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4be6d9c4d55163a755e7c98c0498423b311f38d78a80bbfda600a31680c7e044
+size 442568201

special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

tf_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d6882a73fbb569b938ac3896050b7679a00158bc01984cd7d7b67ce7e1024fe
+size 442752368

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": false, "model_max_length": 512, "special_tokens_map_file": null, "name_or_path": "Maltehb/danish-bert-botxo", "do_basic_tokenize": true, "never_split": null}