Document Classification Script

This script demo.py classifies a given document into a category, identifies relevant narratives, and determines sub-narratives using a fine-tuned language model.

Usage

python demo.py <document_path> <model_path>

• <document_path>: Path to the text file containing the document.
• <model_path>: Path to the fine-tuned model directory.

Output Format

The script prints the results in the following format:

Filename: <document_path>
Category: <Category>
Narratives (Format - Category: Main Narrative: Sub-Narrative):
 - <Category>: <Main Narrative>: <Sub-Narrative>
 - ...
-------------------------

For example:

Filename: sample.txt
Category: Ukraine-Russia War
Narratives (Format - Category: Main Narrative: Sub-Narrative):
 - CC: Criticism of climate movement: Climate movement is corrupt
 - CC: Criticism of climate movement: Climate movement is alarmist
 - CC: Questioning the measurements and science: Scientific community is unreliable
-------------------------

Taxonomy details here

Classification Process

1. Category Classification: Determines if the document belongs to:

   • "Ukraine-Russia War" (URW)
   • "Climate Change" (CC)
   • "Other" (if no relevant category applies)

2. Main Narrative Identification: Selects the most relevant narrative(s) based on predefined categories.

3. Sub-Narrative Identification: Further classifies into specific sub-narratives.

Dependencies

• Python 3.x
• unsloth for model inference
• JSON files in Dataset/ for narrative mappings