| --- |
| license: mit |
| language: |
| - ko |
| library_name: transformers |
| base_model: |
| - beomi/KcELECTRA-base |
| model_type: electra |
| pipeline_tag: text-classification |
| tags: |
| - korean |
| - text-classification |
| - multi-label-classification |
| - electra |
| - kcelectra |
| - fine-tuned |
| datasets: |
| - suicide_related_news_comments_ko |
| metrics: |
| - micro-f1 |
| - macro-f1 |
| - subset-accuracy |
| task_categories: |
| - text-classification |
| task_ids: |
| - multi-label-classification |
| model-index: |
| - name: KcELECTRA-base-finetuned-suicide-comments |
| results: |
| - task: |
| type: multi-label text classification |
| name: Multi-label Text Classification |
| dataset: |
| name: suicide_related_news_comments_ko |
| type: custom |
| split: "8:1:1 (train 7119 / val 890 / test 890)" |
| metrics: |
| - type: micro-f1 |
| value: 0.896 |
| - type: macro-f1 |
| value: 0.892 |
| - type: subset-accuracy |
| value: 0.516 |
| --- |
| |
| # KcELECTRA-base-finetuned-suicide-comments |
|
|
| # Training Details |
| - Data Split: 8:1:1 (Train: 7,119 / Validation: 890 / Test: 890) |
| - Tokenizer: SentencePiece (KcELECTRA tokenizer) |
| - Max Length: 256 |
| - Learning Rate: 3e-5 |
| - Batch Size: 16 |
| - Epochs: 6 |
| - Early Stopping: Patience = 2 |
| - Optimizer: AdamW |
| - Threshold Optimization: Independent per-label tuning (criteria = Micro-F1, Macro-F1) |
| - Thresholds: [0.25, 0.675, 0.8, 0.75, 0.7] |
|
|
| --- |
|
|
| # Result |
|
|
| | Metric | Value | |
| |--------|--------| |
| | Micro-F1 | 0.896 | |
| | Macro-F1 | 0.892 | |
| | Subset Accuracy | 0.516 | |
|
|
| F1-score: |
| | Emotion | F1-score | |
| |----------|-----------| |
| | Dislike | 0.88 | |
| | Sympathy | 0.90 | |
| | Sadness | 0.684| |
| | Surprised | 0.92 | |
| | Angry | 0.93 | |
|
|
| --- |
|
|
