first release of the fear target model

README.md

@@ -0,0 +1,52 @@
+---
+language: sv
+license: mit
+---
+
+## Swedish BERT models for sentiment analysis, Sentiment targets. 
+[Recorded Future](https://www.recordedfuture.com/) together with [AI Sweden](https://www.ai.se/en) releases two language models for target/role assignment in Swedish. The two models are based on the [KB/bert-base-swedish-cased](https://huggingface.co/KB/bert-base-swedish-cased), the models as has been fine tuned to solve a Named Entety Recognition(NER) token classification task.
+
+This is a downstream model to be used in conjunction with the [Swedish violence sentiment classifier](https://huggingface.co/RecordedFuture/Swedish-Sentiment-Violence) or [Swedish violence sentiment classifier](https://huggingface.co/RecordedFuture/Swedish-Sentiment-Fear). The models are trained to tag parts of sentences that has recieved a positive classification from the upstream sentiment classifier. The model will tag parts of sentences that contains the targets that the upstream model has activated on. 
+
+The NER sentiment target models do work as standalone models but their recommended application is downstreamfrom a sentence classification model.  
+
+The models are only trained on Swedish data and only supports inference of Swedish input texts. The models inference metrics for all non-Swedish inputs are not defined, these inputs are considered as out of domain data.
+
+The current models are supported at Transformers version >= 4.3.3 and Torch version 1.8.0, compatibility with older versions are not verified. 
+
+### Fear targets
+
+The model can be imported from the transformers library by running
+
+    from transformers import BertForSequenceClassification, BertTokenizerFast
+    
+    tokenizer = BertTokenizerFast.from_pretrained("RecordedFuture/Swedish-Sentiment-Fear-Targets")
+    classifier_fear_targets= BertForTokenClassification.from_pretrained("RecordedFuture/Swedish-Sentiment-Fear-Targets") 
+
+When the model and tokenizer are initialized the model can be used for inference. 
+
+#### Verification metrics 
+
+During training the Fear target model had the following verification metrics when using "any overlap" as the evaluation metric.  
+
+
+| F-score | Precision | Recall |
+|:-------------------------:|:-------:|:---------:|:------:|
+|  0.8361 |   0.7903  | 0.8876 |
+
+#### Swedish-Sentiment-Violence
+The model be can imported from the transformers library by running
+
+    from transformers import BertForSequenceClassification, BertTokenizerFast
+    
+    tokenizer = BertTokenizerFast.from_pretrained("RecordedFuture/Swedish-Sentiment-Violence-Targets")
+    classifier_violence_targets = BertForTokenClassification.from_pretrained("RecordedFuture/Swedish-Sentiment-Violence-Targets") 
+
+When the model and tokenizer are initialized the model can be used for inference. 
+
+#### Verification metrics 
+During training the Violence target model had the following verification metrics when using "any overlap" as the evaluation metric.  
+
+| F-score | Precision | Recall |
+|:-------------------------:|:-------:|:---------:|:------:|
+|  0.7831|   0.9155|  0.8442 |

config.json

@@ -0,0 +1,36 @@
+{
+  "_name_or_path": "RecordedFuture/Swedish-Sentiment-Fear-Targets",
+  "architectures": [
+    "BertForTokenClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "finetuning_task": "ner",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2
+  },
+  "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,
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.5.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 50325
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebe777c682a3a598a057bbb64f0c52b391b86a2b9d2b20e5392a0fdb6a35dc1e
+size 496497168

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 sha256:4c97f63693224faa166de0721ed1d2024098fb92579ea3ba7d4f471adbb318ff
+size 496679008

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": false, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": false, "special_tokens_map_file": "/home/fmoller/.cache/huggingface/transformers/37f2eab7cd9b3716ce0160ea9562138ae9247fb3ea61a2fd0190b16d0970444e.dd8bd9bfd3664b530ea4e645105f557769387b3da9f79bdb55ed556bdd80611d", "name_or_path": "KB/bert-base-swedish-cased", "do_basic_tokenize": true, "never_split": null}

trainer_state.json

@@ -0,0 +1,133 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "global_step": 320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8748680042238648,
+      "eval_f1": 0.0,
+      "eval_loss": 0.616299569606781,
+      "eval_precision": 0.0,
+      "eval_recall": 0.0,
+      "eval_runtime": 0.1116,
+      "eval_samples_per_second": 815.287,
+      "step": 32
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8785638859556494,
+      "eval_f1": 0.0,
+      "eval_loss": 0.4105246365070343,
+      "eval_precision": 0.0,
+      "eval_recall": 0.0,
+      "eval_runtime": 0.1118,
+      "eval_samples_per_second": 814.082,
+      "step": 64
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.883843717001056,
+      "eval_f1": 0.3583333333333333,
+      "eval_loss": 0.328654408454895,
+      "eval_precision": 0.4387755102040816,
+      "eval_recall": 0.3028169014084507,
+      "eval_runtime": 0.1118,
+      "eval_samples_per_second": 814.03,
+      "step": 96
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.8912354804646251,
+      "eval_f1": 0.36909871244635195,
+      "eval_loss": 0.31546750664711,
+      "eval_precision": 0.4725274725274725,
+      "eval_recall": 0.3028169014084507,
+      "eval_runtime": 0.1118,
+      "eval_samples_per_second": 813.707,
+      "step": 128
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.8933474128827877,
+      "eval_f1": 0.41538461538461535,
+      "eval_loss": 0.3068830370903015,
+      "eval_precision": 0.4576271186440678,
+      "eval_recall": 0.38028169014084506,
+      "eval_runtime": 0.1247,
+      "eval_samples_per_second": 730.008,
+      "step": 160
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.8912354804646251,
+      "eval_f1": 0.48135593220338985,
+      "eval_loss": 0.330695241689682,
+      "eval_precision": 0.46405228758169936,
+      "eval_recall": 0.5,
+      "eval_runtime": 0.1127,
+      "eval_samples_per_second": 807.78,
+      "step": 192
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.895987328405491,
+      "eval_f1": 0.4470588235294118,
+      "eval_loss": 0.3800097107887268,
+      "eval_precision": 0.504424778761062,
+      "eval_recall": 0.4014084507042254,
+      "eval_runtime": 0.1125,
+      "eval_samples_per_second": 808.798,
+      "step": 224
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.899155227032735,
+      "eval_f1": 0.49295774647887325,
+      "eval_loss": 0.4225572347640991,
+      "eval_precision": 0.49295774647887325,
+      "eval_recall": 0.49295774647887325,
+      "eval_runtime": 0.1126,
+      "eval_samples_per_second": 808.356,
+      "step": 256
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.8922914466737064,
+      "eval_f1": 0.4901960784313726,
+      "eval_loss": 0.4346790611743927,
+      "eval_precision": 0.4573170731707317,
+      "eval_recall": 0.528169014084507,
+      "eval_runtime": 0.1123,
+      "eval_samples_per_second": 810.599,
+      "step": 288
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.8870116156282999,
+      "eval_f1": 0.47647058823529415,
+      "eval_loss": 0.48350322246551514,
+      "eval_precision": 0.4090909090909091,
+      "eval_recall": 0.5704225352112676,
+      "eval_runtime": 0.1129,
+      "eval_samples_per_second": 805.916,
+      "step": 320
+    },
+    {
+      "epoch": 10.0,
+      "step": 320,
+      "total_flos": 201896658999468.0,
+      "train_runtime": 25.0952,
+      "train_samples_per_second": 12.751
+    }
+  ],
+  "max_steps": 320,
+  "num_train_epochs": 10,
+  "total_flos": 201896658999468.0,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5efba65dce7e58ef6178bdbdcc23181aa0f6093ee3c26463e169b74b62b6ea48
+size 2351