First version of the da-bert-emotion-classification model and tokenizer

README.md

@@ -0,0 +1,47 @@
+---
+language:
+- da
+tags:
+- bert
+- pytorch
+- emotion
+license: CC-BY_4.0
+datasets:
+- social media
+metrics:
+- f1
+widget:
+- text: "Jeg ejer en rød bil og det er en god bil."
+---
+
+# Danish BERT for emotion classification
+
+The BERT Emotion model classifies a Danish text in one of the following class:
+* Glæde/Sindsro
+* Tillid/Accept
+* Forventning/Interrese
+* Overasket/Målløs
+* Vrede/Irritation
+* Foragt/Modvilje
+* Sorg/trist
+* Frygt/Bekymret
+
+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. 
+
+This model should be used after detecting whether the text contains emotion or not, using the binary [BERT Emotion model](https://huggingface.co/DaNLP/da-bert-emotion-binary).
+
+See the [DaNLP documentation](https://danlp-alexandra.readthedocs.io/en/latest/docs/tasks/sentiment_analysis.html#bert-emotion) for more details. 
+
+Here is how to use the model:
+
+```python
+from transformers import BertTokenizer, BertForSequenceClassification
+
+model = BertForSequenceClassification.from_pretrained("DaNLP/da-bert-emotion-classification")
+tokenizer = BertTokenizer.from_pretrained("DaNLP/da-bert-emotion-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.
+

added_tokens.json

@@ -0,0 +1 @@
+{}

config.json

@@ -0,0 +1,49 @@
+{
+  "_name_or_path": ".",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "directionality": "bidi",
+  "finetuning_task": "emo",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "Gl\u00e6de/Sindsro",
+    "1": "Tillid/Accept",
+    "2": "Forventning/Interrese",
+    "3": "Overasket/M\u00e5ll\u00f8s",
+    "4": "Vrede/Irritation",
+    "5": "Foragt/Modvilje",
+    "6": "Sorg/trist",
+    "7": "Frygt/Bekymret"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "Foragt/Modvilje": "5",
+    "Forventning/Interrese": "2",
+    "Frygt/Bekymret": "7",
+    "Gl\u00e6de/Sindsro": "0",
+    "Overasket/M\u00e5ll\u00f8s": "3",
+    "Sorg/trist": "6",
+    "Tillid/Accept": "1",
+    "Vrede/Irritation": "4"
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "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",
+  "type_vocab_size": 2,
+  "vocab_size": 32000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1913b1dd404d8358e9b42444fb83339be802af550fa749a2cb82f73702941a8f
+size 442580489

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 sha256:473a893298f80a1f7573d57d4df6ae0f78128c8205fb10ab3c85ad5792293619
+size 442764672

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "init_inputs": []}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8914ed596080bcc449b7e1ec71dc4cf483350fc108917ec2d757b826c72bdfcd
+size 1257