README.md

metadata

language: vi
tags:
  - spam-detection
  - vietnamese
  - phobert
license: apache-2.0
datasets:
  - visolex/ViSpamReviews
metrics:
  - accuracy
  - f1
model-index:
  - name: phobert-spam-classification
    results:
      - task:
          type: text-classification
          name: Spam Detection (Multi-Class)
        dataset:
          name: ViSpamReviews
          type: custom
        metrics:
          - name: Accuracy
            type: accuracy
            value: <INSERT_ACCURACY>
          - name: F1 Score
            type: f1
            value: <INSERT_F1_SCORE>
base_model:
  - vinai/phobert-base
pipeline_tag: text-classification

PhoBERT-Spam-MultiClass

Fine-tuned from vinai/phobert-base on ViSpamReviews (multi-class).

• Task: 4-way classification

• Dataset: ViSpamReviews

• Hyperparameters

  • Batch size: 32
  • LR: 3e-5
  • Epochs: 100
  • Max seq len: 256

Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("visolex/phobert-spam-classification")
model = AutoModelForSequenceClassification.from_pretrained("visolex/phobert-spam-classification")

text = "Chỉ PR thương hiệu chứ không review thật."
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)
pred = model(**inputs).logits.argmax(dim=-1).item()
label_map = {0: "NO-SPAM",1: "SPAM-1",2: "SPAM-2",3: "SPAM-3"}
print(label_map[pred])