This model is a fine-tuned BERT-based classifier designed to detect media framing, following Entman's framing theory. The model classifies news texts based on predefined framing categories, enabling researchers to conduct computational framing analysis.

Model Details

Model Description

Developed by: Nurdyansa
Model type: BERT-based text classification model
Language(s): English
License: MIT
Finetuned from model: bert-base-uncased

Framing-BERT: Multi-label Classification for Framing Elements Detection

This repository provides a fine-tuned BERT-based model for detecting framing elements in text, particularly based on the four key elements identified by Entman (1993). The model is designed to classify multiple framing elements simultaneously using multi-label classification.

Framing Elements

The model predicts the presence of the following framing elements, derived from Entman's framing theory:

define_problem – Indicates whether the text defines a social or political problem.
diagnose_cause – Detects if the text attributes causes or sources for the issue.
moral_judgment – Identifies normative or value-laden evaluations.
suggest_remedy – Indicates whether the text proposes solutions, actions, or remedies.

These categories correspond to Entman’s (1993) core functions of framing: defining problems, diagnosing causes, making moral judgments, and suggesting remedies.

Model Information

Base model: bert-base-uncased
Fine-tuned task: Multi-label text classification
Number of labels: 4
Best performance (F1 Macro): 0.6386

Training Details

Optimizer: AdamW
Scheduler: Linear with warmup
Hyperparameter search: Optuna
Best hyperparameters:
- learning_rate: 4.24e-05
- weight_decay: 0.222
- num_train_epochs: 3

Metrics per Epoch (Best Run)

Epoch	Train Loss	Val Loss	Accuracy	F1 Macro	Precision Macro	Recall Macro
1	0.6594	0.6530	0.220	0.6758	0.5775	0.8178
2	0.6027	0.6355	0.2317	0.6329	0.6204	0.6490
3	0.5577	0.6404	0.2283	0.6386	0.6280	0.6566

Example Prediction

python

Predicted Labels:
{
  "define_problem": True,
  "diagnose_cause": True,
  "moral_judgment": True,
  "suggest_remedy": True
}

How to Use

python

from transformers import BertTokenizerFast, BertForSequenceClassification
import torch

tokenizer = BertTokenizerFast.from_pretrained("nurdyansa/framing-bert-model")
model = BertForSequenceClassification.from_pretrained("nurdyansa/framing-bert-model")

text = "Government policies have led to an increase in unemployment, which is unacceptable and must be addressed through immediate reform."

inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
with torch.no_grad():
    logits = model(**inputs).logits
predictions = torch.sigmoid(logits).squeeze().tolist()

labels = ["define_problem", "diagnose_cause", "moral_judgment", "suggest_remedy"]
predicted_labels = {label: pred > 0.5 for label, pred in zip(labels, predictions)}
print(predicted_labels)

Citation

If you use this model in your research or application, please cite it as:

bibtex

@misc{nurdyansa_2025,
	author       = { Nurdyansa },
	title        = { framing-bert-model (Revision f03db73) },
	year         = 2025,
	url          = { https://huggingface.co/nurdyansa/framing-bert-model },
	doi          = { 10.57967/hf/5387 },
	publisher    = { Hugging Face }
}

Contributing

I'm very welcome to invite researchers and practitioners to collaborate in enhancing this model’s precision. Please contribute by:

Providing more annotated data.
Improving label consistency or adding nuance.
Suggesting improvements to model architecture or training methods.

If you are interested in collaborating, sharing insights, or further developing this model, feel free to reach out:

📧 Email: nurdyansa@gmail.com