Persian-Emotion-LLM-Labeler

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

Persian-Emotion-LLM-Labeler

(★ 14)

High-precision Persian emotion annotation pipeline using OpenRouter API (Llama 3.3 70B). Features advanced text normalization, Chain-of-Thought (CoT) reasoning, and fault-tolerant batch processing. Topics (Tags):

Persian-Emotion-LLM-Labeler Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Persian Emotion Labeling with LLMs 🧠🇮🇷

## Overview
This project implements an automated, high-precision text annotation pipeline for classifying Persian social media texts into **11 distinct emotional categories**.

By leveraging the **OpenRouter API**, this system utilizes the state-of-the-art **Meta Llama 3.3 70B Instruct** model. This architecture overcomes local hardware limitations (VRAM) and provides human-level accuracy in understanding Persian nuances, sarcasm, and idioms.

## Features 🚀
- **SOTA Model:** Uses `meta-llama/llama-3.3-70b-instruct` for deep reasoning.
- **Robustness:** Includes Auto-Save and Resume capabilities. If the internet cuts off, the script resumes exactly where it left off.
- **Preprocessing:** Advanced normalization for Persian text (fixing `ی`/`ک`, removing URLs/mentions).
- **Reasoning:** Implements Chain-of-Thought (CoT) prompting to ensure the model explains its decision before labeling.

## Project Structure 📂

```text
/
├── data/
│   ├── input/
│   │   └── Emotions-data.xlsx    <-- Your source file
│   └── output/
│       └── Emotions_Labeled_OpenRouter.xlsx <-- Generated results
│
├── notebooks/
│   └── emotion_labeler.ipynb     <-- Main Python script
│
├── README.md
└── requirements.txt
```

## Setup & Usage 🛠️

### 1. Install Dependencies
Open your terminal in the project root and run:
```bash
pip install -r requirements.txt
```

### 2. Configure API Key
Open `notebooks/emotion_labeler.ipynb` and replace the placeholder with your OpenRouter API key:
```python
OPENROUTER_API_KEY = "sk-or-v1-..."
```

### 3. Run the Labeler
Open the notebook in VS Code and run all cells.
- The script automatically detects the input file in `data/input`.
- It saves progress every 5 records to `data/output`.

## Methodology 🧠

The system uses a **System Prompt** designed for a "Persian Psychologist Persona". It strictly follows these definitions:
1. **Happy** (خوشحال)
2. **Sad** (ناراحت)
3. **Fear** (ترس)
4. **Anger** (خشم)
5. **Surprise** (تعجب)
6. **Disgust** (تنفر)
7. **Shame** (شرم)
8. **Interest** (علاقه)
9. **Envy** (حسادت)
10. **Calmness** (آرامش)
11. **Neutral** (خنثی)

The output is forced into a valid **JSON** format containing:
- `reasoning`: Why the model chose the label.
- `label`: The selected emotion.
- `confidence`: A score from 0-100.

Persian-Emotion-LLM-Labeler

Persian-Emotion-LLM-Labeler Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Persian-Emotion-LLM-Labeler

Persian-Emotion-LLM-Labeler Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README