---
language: en
license: mit
library_name: pytorch
tags:
- climate-journalism
- cjis-copilot
- themes
- multi-label-classification
# Base: Custom MLM-trained DeBERTa (not on HuggingFace Hub)
---

# CJIS Co-pilot: Themes Classifier

Part of the Climate Journalism Intelligence System (CJIS) v3.1 "Journalistic Co-pilot".

## Architecture
- Base: Climate-pretrained DeBERTa (MLM on 20k climate articles)
- Head: Specialist classifier for themes

## Task
Themes classification for climate journalism articles.

## Labels (10)
- `emissions_carbon_pricing`
- `climate_finance_development`
- `food_security_public_health`
- `renewable_energy_tech`
- `fossil_fuel_industry`
- `grid_infrastructure`
- `just_transition`
- `climate_litigation`
- `extreme_weather_events`
- `biodiversity_ecosystems`

## Usage
```python
from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="maai-org/cjis-themes-classifier",
    return_all_scores=True
)

result = classifier("Your climate article here...")