Propaganda Techniques Analysis en B E R T

Introduction

The Propaganda Techniques Analysis BERT model is designed to predict propaganda techniques in English news articles. This model leverages a BERT architecture to perform fine-grained analysis by detecting specific propaganda techniques within texts.

Architecture

The model utilizes a BERT-based architecture, specifically designed for token and sequence joint classification tasks. It classifies text sequences and tokens to identify propaganda techniques, leveraging a pre-trained BERT model (bert-base-cased) as its backbone.

Training

The model was trained using a dataset of news articles annotated at the fragment level with 18 different propaganda techniques. This approach allows for a detailed analysis of texts, identifying not only the presence of propaganda but also categorizing the specific techniques used.

Guide: Running Locally

To run the model locally, follow these steps:

1. Install Dependencies:

   • Ensure you have Python installed.
   • Install the Hugging Face Transformers library:

     pip install transformers
     

2. Load the Model:

   • Use the following Python script to load and run the model:

     from transformers import BertTokenizerFast
     from .model import BertForTokenAndSequenceJointClassification
     
     tokenizer = BertTokenizerFast.from_pretrained('bert-base-cased')
     model = BertForTokenAndSequenceJointClassification.from_pretrained(
         "QCRI/PropagandaTechniquesAnalysis-en-BERT",
         revision="v0.1.0",
     )
     
     inputs = tokenizer.encode_plus("Hello, my dog is cute", return_tensors="pt")
     outputs = model(**inputs)
     sequence_class_index = torch.argmax(outputs.sequence_logits, dim=-1)
     sequence_class = model.sequence_tags[sequence_class_index[0]]
     token_class_index = torch.argmax(outputs.token_logits, dim=-1)
     tokens = tokenizer.convert_ids_to_tokens(inputs.input_ids[0][1:-1])
     tags = [model.token_tags[i] for i in token_class_index[0].tolist()[1:-1]]
     

3. Suggest Cloud GPUs:

   • For performance improvements, consider using cloud GPUs such as those offered by AWS, Google Cloud, or Azure.

License

The model is licensed under the MIT License, allowing for wide use and distribution.